Stata Annotated Output
Regression Analysis
This page shows an example regression analysis with footnotes explaining the output. These data were collected on 200 high schools students and are scores on various tests, including science, math, reading and social studies (socst). The variable female is a dichotomous variable coded 1 if the student was female and 0 if male.
use http://www.ats.ucla.edu/stat/stata/notes/hsb2
(highschool and beyond (200 cases))
regress science math female socst read
Source | SS df MS Number of obs = 200
-------------+------------------------------ F( 4, 195) = 46.69
Model | 9543.72074 4 2385.93019 Prob > F = 0.0000
Residual | 9963.77926 195 51.0963039 R-squared = 0.4892
-------------+------------------------------ Adj R-squared = 0.4788
Total | 19507.5 199 98.0276382 Root MSE = 7.1482
------------------------------------------------------------------------------
science | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
math | .3893102 .0741243 5.25 0.000 .243122 .5354983
female | -2.009765 1.022717 -1.97 0.051 -4.026772 .0072428
socst | .0498443 .062232 0.80 0.424 -.0728899 .1725784
read | .3352998 .0727788 4.61 0.000 .1917651 .4788345
_cons | 12.32529 3.193557 3.86 0.000 6.026943 18.62364
------------------------------------------------------------------------------
________________________________________
Anova Table
Sourcea | SSb dfc MSd
-------------+------------------------------
Model | 9543.72074 4 2385.93019
Residual | 9963.77926 195 51.0963039
-------------+------------------------------
Total | 19507.5 199 98.0276382
a. Source - This is the source of variance, Model, Residual, and Total. The Total variance is partitioned into the variance which can be explained by the independent variables (Model) and the variance which is not explained by the independent variables (Residual, sometimes called Error). Note that the Sums of Squares for the Model and Residual add up to the Total Variance, reflecting the fact that the Total Variance is partitioned into Model and Residual variance.
b. SS - These are the Sum of Squares associated with the three sources of variance, Total, Model and Residual. These can be computed in many ways. Conceptually, these formulas can be expressed as:
SSTotal The total variability around the mean. S(Y - Ybar)2.
SSResidual The sum of squared errors in prediction. S(Y - Ypredicted)2.
SSModel The improvement in prediction by using the predicted value of Y over just using the mean of Y. Hence, this would be the squared differences between the predicted value of Y and the mean of Y, S(Ypredicted - Ybar)2. Another way to think of this is the SSModel is SSTotal - SSResidual. Note that the SSTotal = SSModel + SSResidual. Note that SSModel / SSTotal is equal to .4892, the value of R-Square. This is because R-Square is the proportion of the variance explained by the independent variables, hence can be computed by SSModel / SSTotal.
c. df - These are the degrees of freedom associated with the sources of variance. The total variance has N-1 degrees of freedom. In this case, there were N=200 students, so the DF for total is 199. The model degrees of freedom corresponds to the number of predictors minus 1 (K-1). You may think this would be 4-1 (since there were 4 independent variables in the model, math, female, socst and read). But, the intercept is automatically included in the model (unless you explicitly omit the intercept). Including the intercept, there are 5 predictors, so the model has 5-1=4 degrees of freedom. The Residual degrees of freedom is the DF total minus the DF model, 199 - 4 is 195.
d. MS - These are the Mean Squares, the Sum of Squares divided by their respective DF. For the Model, 9543.72074 / 4 = 2385.93019. For the Residual, 9963.77926 / 195 = 51.0963039. These are computed so you can compute the F ratio, dividing the Mean Square Model by the Mean Square Residual to test the significance of the predictors in the model.
________________________________________
Overall Model Fit
Number of obse = 200
F( 4, 195)f = 46.69
Prob > Ff = 0.0000
R-squaredg = 0.4892
Adj R-squaredh = 0.4788
Root MSEi = 7.1482
e. Number of obs - This is the number of observations used in the regression analysis.
f. F and Prob > F - The F-value is the Mean Square Model (2385.93019) divided by the Mean Square Residual (51.0963039), yielding F=46.69. The p-value associated with this F value is very small (0.0000). These values are used to answer the question "Do the independent variables reliably predict the dependent variable?". The p-value is compared to your alpha level (typically 0.05) and, if smaller, you can conclude "Yes, the independent variables reliably predict the dependent variable". You could say that the group of variables math and female can be used to reliably predict write (the dependent variable). If the p-value were greater than 0.05, you would say that the group of independent variables does not show a statistically significant relationship with the dependent variable, or that the group of independent variables does not reliably predict the dependent variable. Note that this is an overall significance test assessing whether the group of independent variables when used together reliably predict the dependent variable, and does not address the ability of any of the particular independent variables to predict the dependent variable. The ability of each individual independent variable to predict the dependent variable is addressed in the table below where each of the individual variables are listed.
g. R-squared - R-Squared is the proportion of variance in the dependent variable (science) which can be predicted from the independent variables (math, female, socst and read). This value indicates that 48.92% of the variance in science scores can be predicted from the variables math, female, socst and read. Note that this is an overall measure of the strength of association, and does not reflect the extent to which any particular independent variable is associated with the dependent variable.
h. Adj R-squared - Adjusted R-square. As predictors are added to the model, each predictor will explain some of the variance in the dependent variable simply due to chance. One could continue to add predictors to the model which would continue to improve the ability of the predictors to explain the dependent variable, although some of this increase in R-square would be simply due to chance variation in that particular sample. The adjusted R-square attempts to yield a more honest value to estimate the R-squared for the population. The value of R-square was .4892, while the value of Adjusted R-square was .4788 Adjusted R-squared is computed using the formula 1 - ((1 - Rsq)((N - 1) /( N - k - 1)). From this formula, you can see that when the number of observations is small and the number of predictors is large, there will be a much greater difference between R-square and adjusted R-square (because the ratio of (N - 1) / (N - k - 1) will be much greater than 1). By contrast, when the number of observations is very large compared to the number of predictors, the value of R-square and adjusted R-square will be much closer because the ratio of (N - 1)/(N - k - 1) will approach 1.
i. Root MSE - Root MSE is the standard deviation of the error term, and is the square root of the Mean Square Residual (or Error).
________________________________________
Parameter Estimates
------------------------------------------------------------------------------
sciencej | Coef.k Std. Err.l tm P>|t|m [95% Conf. Interval]n
-------------+----------------------------------------------------------------
math | .3893102 .0741243 5.25 0.000 .243122 .5354983
female | -2.009765 1.022717 -1.97 0.051 -4.026772 .0072428
socst | .0498443 .062232 0.80 0.424 -.0728899 .1725784
read | .3352998 .0727788 4.61 0.000 .1917651 .4788345
_cons | 12.32529 3.193557 3.86 0.000 6.026943 18.62364
------------------------------------------------------------------------------
j. science - This column shows the dependent variable at the top (science) with the predictor variables below it (math, female, socst, read and _cons). The last variable (_cons) represents the constant, also referred to in textbooks as the Y intercept, the height of the regression line when it crosses the Y axis. In other words, this is the predicted value of science when all other variables are 0.
k. Coef. - These are the values for the regression equation for predicting the dependent variable from the independent variable. The regression equation is presented in many different ways, for example:
Ypredicted = b0 + b1*x1 + b2*x2 + b3*x3 + b4*x4
The column of estimates (coefficients or parameter estimates, from here on labeled coefficients) provides the values for b0, b1, b2, b3 and b4 for this equation. Expressed in terms of the variables used in this example, the regression equation is
sciencePredicted = 12.32529 + .3893102*math + -2.009765*female+.0498443*socst+.3352998*read
These estimates tell you about the relationship between the independent variables and the dependent variable. These estimates tell the amount of increase in science scores that would be predicted by a 1 unit increase in the predictor. Note: For the independent variables which are not significant, the coefficients are not significantly different from 0, which should be taken into account when interpreting the coefficients. (See the columns with the t-value and p-value about testing whether the coefficients are significant).
math - The coefficient (parameter estimate) is .3893102. So, for every unit (i.e., point, since this is the metric in which the tests are measured) increase in math, a .3893102 unit increase in science is predicted, holding all other variables constant. (It does not matter at what value you hold the other variables constant, because it is a linear model.) Or, for every increase of one point on the math test, your science score is predicted to be higher by .3893102 points. This is significantly different from 0.
female - For every unit increase in female, there is a -2.009765 unit decrease in the predicted science score, holding all other variables constant. Since female is coded 0/1 (0=male, 1=female) the interpretation can be put more simply. For females the predicted science score would be 2 points lower than for males. The variable female is technically not statistically significantly different from 0, because the p-value is greater than .05. However, .051 is so close to .05 that some researchers would still consider it to be statistically significant.
socst - The coefficient for socst is .0498443. This means that for a 1-unit increase in the social studies score, we expect an approximately .05 point increase in the science score. This is not statistically significant; in other words, .0498443 is not different from 0.
read - The coefficient for read is .3352998. Hence, for every unit increase in reading score we expect a .34 point increase in the science score. This is statistically significant.
l. Std. Err. - These are the standard errors associated with the coefficients. The standard error is used for testing whether the parameter is significantly different from 0 by dividing the parameter estimate by the standard error to obtain a t-value (see the column with t-values and p-values). The standard errors can also be used to form a confidence interval for the parameter, as shown in the last two columns of this table.
m. t and P>|t| - These columns provide the t-value and 2-tailed p-value used in testing the null hypothesis that the coefficient (parameter) is 0. If you use a 2-tailed test, then you would compare each p-value to your preselected value of alpha. Coefficients having p-values less than alpha are statistically significant. For example, if you chose alpha to be 0.05, coefficients having a p-value of 0.05 or less would be statistically significant (i.e., you can reject the null hypothesis and say that the coefficient is significantly different from 0). If you use a 1-tailed test (i.e., you predict that the parameter will go in a particular direction), then you can divide the p-value by 2 before comparing it to your preselected alpha level. With a 2-tailed test and alpha of 0.05, you may reject the null hypothesis that the coefficient for female is equal to 0. The coefficient of -2.009765 is significantly greater than 0. However, if you used a 2-tailed test and alpha of 0.01, the p-value of .0255 is greater than 0.01 and the coefficient for female would not be significant at the 0.01 level. Had you predicted that this coefficient would be positive (i.e., a one tail test), you would be able to divide the p-value by 2 before comparing it to alpha. This would yield a 1-tailed p-value of 0.00945, which is less than 0.01 and then you could conclude that this coefficient is greater than 0 with a one tailed alpha of 0.01.
The coefficient for math is significantly different from 0 using alpha of 0.05 because its p-value is 0.000, which is smaller than 0.05.
The coefficient for socst (.0498443) is not statistically significantly different from 0 because its p-value is definitely larger than 0.05.
The coefficient for read (.3352998) is statistically significant because its p-value of 0.000 is less than .05.
The constant (_cons) is significantly different from 0 at the 0.05 alpha level. However, having a significant intercept is seldom interesting.
n. [95% Conf. Interval] - This shows a 95% confidence interval for the coefficient. This is very useful as it helps you understand how high and how low the actual population value of the parameter might be. The confidence intervals are related to the p-values such that the coefficient will not be statistically significant if the confidence interval includes 0. If you look at the confidence interval for female, you will see that it just includes 0 (-4 to .007). Because .007 is so close to 0, the p-value is close to .05. If the upper confidence level had been a little smaller, such that it did not include 0, the coefficient for female would have been statistically significant. Also, consider the coefficients for female (-2) and read (.34). Immediately you see that the estimate for female is so much bigger, but examine the confidence interval for it (-4 to .007). Now examine the confidence interval for read (.19 to .48). Even though female has a bigger coefficient (in absolute terms) it could be as small as -4. By contrast, the lower confidence level for read is 19, which is still above 0. So, even though female has a bigger coefficient, read is significant and even the smallest value in the confidence interval is still higher than 0. The same cannot be said about the coefficient for socst. Such confidence intervals help you to put the estimate from the coefficient into perspective by seeing how much the value could vary.
Monday, August 10, 2009
Analisis Regresi
ANALISIS REGRESI
Pengantar
Regresi merupakan salah satu model ekonomi, yaitu representasi dari fenomena ekonomi dalam bentuk persamaan matematis.
Regresi merupakan bagian dari ilmu statistika, yang banyak digunakan dalam analisis ekonomi.
Dalam ilmu ekonomi, analisis regresi termasuk dalam ekonometrika, yaitu ilmu yang mempelajari alat untuk mengukur perekonomian (economic measurement)
Analisis Regresi digunakan untuk:
Menduga arah dari hubungan kuantitatif diantara variabel-variabel eknomi.
Menguji hipotesis tentang ekonomi
Meramalkan masa depan perekonomian.
Beberapa hal yang dibutuhkan sebelum melakukan analisis regresi:
1. Penguasaan teori dari permasalahan yang ingin diselesaikan
2. Matematika
3. Teori dasar Statistika
4. Data
5. Perlengkapan untuk komputasi
6. Kemampuan intepretasi berdasarkan pengetahuan
ekonomi/umum (common sense)
Proses Analisis dengan Model
Beberapa hal yang dibutuhkan setelah melakukan analisis regresi:
1. Menganalisis apakah hasil pengukuran sudah tepat.
2. Menyimpulkan hasil pengujian hipotesis.
3. Melakukan peramalan/prediksi
4. Melakukan evaluasi hasil
Definisi Regresi Linear
Regresi linear adalah teknik statistika yang memberikan pendugaan dari kemiringan suatu garis lurus (linear), dan posisi dimana garis tersebut memotong sumbu y, berdasarkan sejumlah informasi mengenai hubungan antar variabel.
Memberikan pendugaan nilai a dan b, berdasarkan sejumlah informasi mengenai x dan y, pada persamaan berikut:
y=a+b.x
x disebut variabel independent, karena nilainya tidak tergantung variabel lain
y disebut variabel dependent, karena nilainya tergantung nilai x
a dan b disebut parameter, a adalah intercept dan b adalah slope.
Regresi linear sederhana, apabila variabel dependent hanya ditentukan oleh satu variabel independent.
y = a + b.x + e
Regresi Linear berganda, apabila variabel dependent ditentukan oleh lebih dari satu variabel independent.
y = a + b1.x1 + b2.x2 + …. + bn.xn + e
e = error term = perbedaan antara y aktual dengan y hasil estimasi garis regresi
Metode yang digunakan untuk mengestimasi parameter garis regresi disebut Metode Kuadrat Terkecil (Least Square Method)
Metode ini meminimisasi jumlah dari error yang dikuadratkan dari setiap observasi.
Rumus untuk slope garis regresi adalah:
Rumus untuk intercept garis regresi adalah:
Asumsi-asumsi klasik regresi linier :
1. Model linear dalam parameter.
2. Nilai x tetap dalam pengambilan sampel yang diulang.
3. Nilai rata-rata dari error sama dengan nol.
4. Homoskedastis yaitu nilai varians dari setiap error sama.
5. Tidak ada korelasi antar error.
6. Covarians antara ui dan xi adalah nol.
7. Banyaknya observasi n harus lebih besar daripada banyaknya parameter yang diestimasi.
8. Nilai dari xi harus bervariasi (tidak boleh sama).
9. Model regresi dispesifikasikan dengan benar.
10. Tidak ada multikolinearitas sempurna.
Aplikasi pada program stata :
Sebelum melakukan regresi sebaiknya dilkaukan pengaturan waktu observasi dengan perintah:
tsset [nama variabel]
Contoh :
tsset obs
perintah ini akan menjadikan variabel obs(observasi) menjadi satuan waktu
Untuk melakukan regersi OLS dapat menggunakan perintah
regress [variabel dependent] [variabel independent]...[variabel independen ke-n]
contoh :
regress x5 x1 x2 x3 x4 mb obs
perintah ini akan memerintahkan stata untuk melakukan regresi linier dengan variabel dependent x5 menggunakan variabel independent x1 x2 x3 x4 mb dan obs.
Evaluasi hasil regresi :
1. Lakukan evaluasi terhadap tanda dari slope, apakah sudah sesuai dengan teori. Jika belum, ada kemungkinan data yang digunakan dan spesifikasi model regresi salah.
2. Lakukan evaluasi terhadap signifikasi dari parameter, yang ditunjukkan oleh nilai t-stat dan prob t-stat. Apabila nilai prob t-stat < tingkat signifikansi (a, biasanya 0.05 atau 0.1) maka variabel independent tersebut secara statistik signifikan mempengaruhi variabel dependent.
3. Lakukan evaluasi terhadap nilai R2 yang menunjukkan persentase variasi dari variabel dependent yang dapat dijelaskan oleh variabel-variabel independent.
4. Lakukan evaluasi terhadap asumsi-asumsi regresi, autokorelasi, multikolinearitas, heteroskedastisitas.
5. Perbaiki model regresi dan data, lakukan kembali estimasi sampai memperoleh hasil yang terbaik.
Pelanggaran Asumsi :
Multikolinearitas
Adanya hubungan linear antara beberapa atau semua variabel penjelas (variabel exogen) dalam model regresi.
Contoh:
Dimana: x1i = 2 x2i
Hasil pendugaan parameter regresi akan menjadi bias
Mencurigai adanya multikolinearitas:
1. Nilai t untuk setiap parameter tidak significant tetapi nilai R2 sangat tinggi (mendekati 1).
2. Korelasi berpasangan diantara variabel penjelas sangat tinggi.
3. Nilai R2 dari auxiliary regression lebih besar dari nilai R2 model yang sebenarnya
Auxialiary regression: regresi dari setiap Xi terhadap X yang lainnya
Aplikasi pada program stata :
Untuk mendeteksi adanya masalah multikolinearitas pada program stata dapat digunakan perintah :
cor [variabel independent]…[variabel independent ke-n]
contoh :
cor x1 x2 x3 x4 mb obs
Perintah ini akan menghitung corelasi antara variabel independent
Lalu lihat nilai koefisien corelasi antar variabel independent jika menunjukan angka 0,75 baik bernilai negative maupun positif maka dapat disimpulkan adanya multikolinearitas
Atau perintah
vif
perintah ini akan menghitung besarnya variance inflation factors (VIFs) untuk semua variabel independent
Lalu kita dapat melihat nilai vif, bukan mean vif-nya, apbila nilai vifnya bernilai ≥ 20
Mengatasi masalah multikolinearitas:
1. Membuang satu atau beberapa variabel yang mengandung bias.
Cara yang dilakukan adalah melihat hasil perhitungan cor atau vif lalu membuang variabel independent yang memiliki korelas atau variabel yang nilai vif melebihi 20, dengan tidak melupakan tingkat signifikansi variabel itu dalam sebuah model.
2. Melakukan transformasi variabel, biasanya dalam bentuk:
a. First difference:
Membuat variabel first difference pada stata dapat mengunakan perintah :
gen [nama variabel first difference]=d.[variabel awal]
contoh :
gen fdx1=d.x1
perintah itu akan membuat suatu variabel baru yang merupakan variabel first difference dari veriabel x1, kemudian variabel yang telah menjadi first difference telah dapat diregresi.
b. Transformasi rasio:
3. Menambah data
Hetroskedastisitas
Hetrosekadastis terjadi jika nilai varians dari setiap eror berbeda atau bervariasi.
Mendeteksi adanya heteroskedastisitas dapat menggunakan white test:
1. Lakukan estimasi dengan menggunakan metode OLS untuk persamaan:
2. Setelah diperoleh , lakukan regresi dengan menggunakan persamaan:
kemudian tentukan R2 nya
1. Ujilah hipotesis :
H0 : homoskedastis
H1 : heteroskedastis
Dengan menggunakan statistik uji: nR2 ~ cdx2.
df = jumalah parameter dikurangi 1.
Melacak gejala hetroskedastis menggunakan program stata dapat menggunakan perintah hettest
Mengatasi heteroskedastisitas:
1. Menggunakan metode GLS (Generalized Least Squares)
2. Transformasi Variabel:
a. dibagi dengan xi
b. dibagi dengan
c. dibagi dengan
d. transformasi logaritma
Untuk mengatasi adanya masalah heteroskedastisitas pada program stata dapat digunakan perintah perhitungan robust, dengan bentuk perintah :
regress [dependen variabel] [independent variabel]...[variabel independent ke-n], robust
contoh :
regress x5 x1 x2 x3 x4 mb obs, robust
perintah robust akan secra otomatis menghilangkan heteroskedastisitas dengan jalan membobotkan dengan robust standar eror. Hasil regresi yang didapat telah dapat dipastikan telah terbebas dari heteroskedastisitas.
Autokorelasi
Autokorelasi suatu keadan dimana terjadi korelasi eror antar periode waktu.
Cara mendeteksi autokorelasi adalah dengan melakukan pengujian Durbin - Watson test:
Untuk n tertentu tesedia nilai dU dan dL pada table
Ujilah hipotesis:
H0 : tidak ada autokorelasi positif
H0*: tidak ada autokorelasi negatif
Bandingkan dengan diagram berikut ini:
Mendeteksi Autokorelasi menggunakan stata, dapat dilakukan dengan perintah :
durbina
Perintah durbina akan melakukan pengujian Durbin - Watson test
bgodfrey
Perintah bgodfrey akan melakuka pengujian Breusch-Godfrey LM test
Mengatasi Autokorelasi
1. Menggunakan metode GLS (Generalized Least Squares)
2. Menggunakan metode First Difference
Untuk stata dapat dikerjakan dengan cara:
Membuat variabel first difference pada stata dapat mengunakan perintah :
gen [nama variabel first difference]=d.[variabel awal]
contoh :
gen fdx1=d.x1
perintah itu akan membuat suatu variabel baru yang merupakan variabel first difference dari veriabel x1, kemudian variabel yang telah menjadi first difference telah dapat diregresi.
Untuk mengatasi adanya masalah autokorelasi pada program stata dapat digunakan perintah perhitungan prais, dengan bentuk perintah :
prais [dependen variabel] [independent variabel]...[variabel independent ke-n]
contoh :
prais x5 x1 x2 x3 x4 mb obs
Perintah prais akan secara otomatis menghilangkan autokorelasi dengan jalan menambahkan variabel autoregresion dengan lag yang ditentukan oleh program stata. Hasil regresi yang didapat telah dapat dipastikan telah terbebas dari autokorelasi.
Stata dapat menghilangkan gejala heterosekdasitas dan autokorelasi secara langsung dengan mengabungkan perintah robust dan prais. Dengan bentu perintah :
prais [variabel dependent] [variabel independent]…[variabel independent ke-n], robust
Contoh ;
prais x5 x1 x2 x3 x4 mb obs, robust
Gabungan perintah prais dan robust akan secara otomatis menghilangkan autokorelasi dan heteroskedastisitas
Peringatan Bagi Praktisi
Tetap gunakan common sense dan teori ekonomi
Mengutamakan relevansi daripada matematis
Mengetahui konteks
Memeriksa data
Tidak boleh memuja kerumitan
Harus memperhatikan dengan seksama hasil
Hati-hati dengan biaya data mining dan waktu
Harus mau kompromi
Harus mengakui adanya sensitivitas
Pengantar
Regresi merupakan salah satu model ekonomi, yaitu representasi dari fenomena ekonomi dalam bentuk persamaan matematis.
Regresi merupakan bagian dari ilmu statistika, yang banyak digunakan dalam analisis ekonomi.
Dalam ilmu ekonomi, analisis regresi termasuk dalam ekonometrika, yaitu ilmu yang mempelajari alat untuk mengukur perekonomian (economic measurement)
Analisis Regresi digunakan untuk:
Menduga arah dari hubungan kuantitatif diantara variabel-variabel eknomi.
Menguji hipotesis tentang ekonomi
Meramalkan masa depan perekonomian.
Beberapa hal yang dibutuhkan sebelum melakukan analisis regresi:
1. Penguasaan teori dari permasalahan yang ingin diselesaikan
2. Matematika
3. Teori dasar Statistika
4. Data
5. Perlengkapan untuk komputasi
6. Kemampuan intepretasi berdasarkan pengetahuan
ekonomi/umum (common sense)
Proses Analisis dengan Model
Beberapa hal yang dibutuhkan setelah melakukan analisis regresi:
1. Menganalisis apakah hasil pengukuran sudah tepat.
2. Menyimpulkan hasil pengujian hipotesis.
3. Melakukan peramalan/prediksi
4. Melakukan evaluasi hasil
Definisi Regresi Linear
Regresi linear adalah teknik statistika yang memberikan pendugaan dari kemiringan suatu garis lurus (linear), dan posisi dimana garis tersebut memotong sumbu y, berdasarkan sejumlah informasi mengenai hubungan antar variabel.
Memberikan pendugaan nilai a dan b, berdasarkan sejumlah informasi mengenai x dan y, pada persamaan berikut:
y=a+b.x
x disebut variabel independent, karena nilainya tidak tergantung variabel lain
y disebut variabel dependent, karena nilainya tergantung nilai x
a dan b disebut parameter, a adalah intercept dan b adalah slope.
Regresi linear sederhana, apabila variabel dependent hanya ditentukan oleh satu variabel independent.
y = a + b.x + e
Regresi Linear berganda, apabila variabel dependent ditentukan oleh lebih dari satu variabel independent.
y = a + b1.x1 + b2.x2 + …. + bn.xn + e
e = error term = perbedaan antara y aktual dengan y hasil estimasi garis regresi
Metode yang digunakan untuk mengestimasi parameter garis regresi disebut Metode Kuadrat Terkecil (Least Square Method)
Metode ini meminimisasi jumlah dari error yang dikuadratkan dari setiap observasi.
Rumus untuk slope garis regresi adalah:
Rumus untuk intercept garis regresi adalah:
Asumsi-asumsi klasik regresi linier :
1. Model linear dalam parameter.
2. Nilai x tetap dalam pengambilan sampel yang diulang.
3. Nilai rata-rata dari error sama dengan nol.
4. Homoskedastis yaitu nilai varians dari setiap error sama.
5. Tidak ada korelasi antar error.
6. Covarians antara ui dan xi adalah nol.
7. Banyaknya observasi n harus lebih besar daripada banyaknya parameter yang diestimasi.
8. Nilai dari xi harus bervariasi (tidak boleh sama).
9. Model regresi dispesifikasikan dengan benar.
10. Tidak ada multikolinearitas sempurna.
Aplikasi pada program stata :
Sebelum melakukan regresi sebaiknya dilkaukan pengaturan waktu observasi dengan perintah:
tsset [nama variabel]
Contoh :
tsset obs
perintah ini akan menjadikan variabel obs(observasi) menjadi satuan waktu
Untuk melakukan regersi OLS dapat menggunakan perintah
regress [variabel dependent] [variabel independent]...[variabel independen ke-n]
contoh :
regress x5 x1 x2 x3 x4 mb obs
perintah ini akan memerintahkan stata untuk melakukan regresi linier dengan variabel dependent x5 menggunakan variabel independent x1 x2 x3 x4 mb dan obs.
Evaluasi hasil regresi :
1. Lakukan evaluasi terhadap tanda dari slope, apakah sudah sesuai dengan teori. Jika belum, ada kemungkinan data yang digunakan dan spesifikasi model regresi salah.
2. Lakukan evaluasi terhadap signifikasi dari parameter, yang ditunjukkan oleh nilai t-stat dan prob t-stat. Apabila nilai prob t-stat < tingkat signifikansi (a, biasanya 0.05 atau 0.1) maka variabel independent tersebut secara statistik signifikan mempengaruhi variabel dependent.
3. Lakukan evaluasi terhadap nilai R2 yang menunjukkan persentase variasi dari variabel dependent yang dapat dijelaskan oleh variabel-variabel independent.
4. Lakukan evaluasi terhadap asumsi-asumsi regresi, autokorelasi, multikolinearitas, heteroskedastisitas.
5. Perbaiki model regresi dan data, lakukan kembali estimasi sampai memperoleh hasil yang terbaik.
Pelanggaran Asumsi :
Multikolinearitas
Adanya hubungan linear antara beberapa atau semua variabel penjelas (variabel exogen) dalam model regresi.
Contoh:
Dimana: x1i = 2 x2i
Hasil pendugaan parameter regresi akan menjadi bias
Mencurigai adanya multikolinearitas:
1. Nilai t untuk setiap parameter tidak significant tetapi nilai R2 sangat tinggi (mendekati 1).
2. Korelasi berpasangan diantara variabel penjelas sangat tinggi.
3. Nilai R2 dari auxiliary regression lebih besar dari nilai R2 model yang sebenarnya
Auxialiary regression: regresi dari setiap Xi terhadap X yang lainnya
Aplikasi pada program stata :
Untuk mendeteksi adanya masalah multikolinearitas pada program stata dapat digunakan perintah :
cor [variabel independent]…[variabel independent ke-n]
contoh :
cor x1 x2 x3 x4 mb obs
Perintah ini akan menghitung corelasi antara variabel independent
Lalu lihat nilai koefisien corelasi antar variabel independent jika menunjukan angka 0,75 baik bernilai negative maupun positif maka dapat disimpulkan adanya multikolinearitas
Atau perintah
vif
perintah ini akan menghitung besarnya variance inflation factors (VIFs) untuk semua variabel independent
Lalu kita dapat melihat nilai vif, bukan mean vif-nya, apbila nilai vifnya bernilai ≥ 20
Mengatasi masalah multikolinearitas:
1. Membuang satu atau beberapa variabel yang mengandung bias.
Cara yang dilakukan adalah melihat hasil perhitungan cor atau vif lalu membuang variabel independent yang memiliki korelas atau variabel yang nilai vif melebihi 20, dengan tidak melupakan tingkat signifikansi variabel itu dalam sebuah model.
2. Melakukan transformasi variabel, biasanya dalam bentuk:
a. First difference:
Membuat variabel first difference pada stata dapat mengunakan perintah :
gen [nama variabel first difference]=d.[variabel awal]
contoh :
gen fdx1=d.x1
perintah itu akan membuat suatu variabel baru yang merupakan variabel first difference dari veriabel x1, kemudian variabel yang telah menjadi first difference telah dapat diregresi.
b. Transformasi rasio:
3. Menambah data
Hetroskedastisitas
Hetrosekadastis terjadi jika nilai varians dari setiap eror berbeda atau bervariasi.
Mendeteksi adanya heteroskedastisitas dapat menggunakan white test:
1. Lakukan estimasi dengan menggunakan metode OLS untuk persamaan:
2. Setelah diperoleh , lakukan regresi dengan menggunakan persamaan:
kemudian tentukan R2 nya
1. Ujilah hipotesis :
H0 : homoskedastis
H1 : heteroskedastis
Dengan menggunakan statistik uji: nR2 ~ cdx2.
df = jumalah parameter dikurangi 1.
Melacak gejala hetroskedastis menggunakan program stata dapat menggunakan perintah hettest
Mengatasi heteroskedastisitas:
1. Menggunakan metode GLS (Generalized Least Squares)
2. Transformasi Variabel:
a. dibagi dengan xi
b. dibagi dengan
c. dibagi dengan
d. transformasi logaritma
Untuk mengatasi adanya masalah heteroskedastisitas pada program stata dapat digunakan perintah perhitungan robust, dengan bentuk perintah :
regress [dependen variabel] [independent variabel]...[variabel independent ke-n], robust
contoh :
regress x5 x1 x2 x3 x4 mb obs, robust
perintah robust akan secra otomatis menghilangkan heteroskedastisitas dengan jalan membobotkan dengan robust standar eror. Hasil regresi yang didapat telah dapat dipastikan telah terbebas dari heteroskedastisitas.
Autokorelasi
Autokorelasi suatu keadan dimana terjadi korelasi eror antar periode waktu.
Cara mendeteksi autokorelasi adalah dengan melakukan pengujian Durbin - Watson test:
Untuk n tertentu tesedia nilai dU dan dL pada table
Ujilah hipotesis:
H0 : tidak ada autokorelasi positif
H0*: tidak ada autokorelasi negatif
Bandingkan dengan diagram berikut ini:
Mendeteksi Autokorelasi menggunakan stata, dapat dilakukan dengan perintah :
durbina
Perintah durbina akan melakukan pengujian Durbin - Watson test
bgodfrey
Perintah bgodfrey akan melakuka pengujian Breusch-Godfrey LM test
Mengatasi Autokorelasi
1. Menggunakan metode GLS (Generalized Least Squares)
2. Menggunakan metode First Difference
Untuk stata dapat dikerjakan dengan cara:
Membuat variabel first difference pada stata dapat mengunakan perintah :
gen [nama variabel first difference]=d.[variabel awal]
contoh :
gen fdx1=d.x1
perintah itu akan membuat suatu variabel baru yang merupakan variabel first difference dari veriabel x1, kemudian variabel yang telah menjadi first difference telah dapat diregresi.
Untuk mengatasi adanya masalah autokorelasi pada program stata dapat digunakan perintah perhitungan prais, dengan bentuk perintah :
prais [dependen variabel] [independent variabel]...[variabel independent ke-n]
contoh :
prais x5 x1 x2 x3 x4 mb obs
Perintah prais akan secara otomatis menghilangkan autokorelasi dengan jalan menambahkan variabel autoregresion dengan lag yang ditentukan oleh program stata. Hasil regresi yang didapat telah dapat dipastikan telah terbebas dari autokorelasi.
Stata dapat menghilangkan gejala heterosekdasitas dan autokorelasi secara langsung dengan mengabungkan perintah robust dan prais. Dengan bentu perintah :
prais [variabel dependent] [variabel independent]…[variabel independent ke-n], robust
Contoh ;
prais x5 x1 x2 x3 x4 mb obs, robust
Gabungan perintah prais dan robust akan secara otomatis menghilangkan autokorelasi dan heteroskedastisitas
Peringatan Bagi Praktisi
Tetap gunakan common sense dan teori ekonomi
Mengutamakan relevansi daripada matematis
Mengetahui konteks
Memeriksa data
Tidak boleh memuja kerumitan
Harus memperhatikan dengan seksama hasil
Hati-hati dengan biaya data mining dan waktu
Harus mau kompromi
Harus mengakui adanya sensitivitas
Stata: Introduction
Introduction
This page is designed to give you a quick look at some basic tasks you may need to do in Stata. What it does cover is:
1. Overview of Stata.
2. Creating and using "log" and "do" files.
3. Data management.
4. File management.
5. Basic data description.
6. Basics in regression analysis.
There are some other resources you may find useful:
1. Stata manuals
2. Stata Technical Bulletin
3. The DSS web site, www.princeton.edu/~data/stata (the major source for this overview)
4. Stata’s web site, www.stata.com
5. Stata’s on-line tutorial (the command, ‘help tutorial’ will show you what to do).
You can click on the to see what the output from that particular command or commands look like, or what a particular window should look like in the PC version of Stata; use the "back" button on your browser to return to this page.
________________________________________
Overview
Stata is an interactive data analysis program which runs on a variety of platforms. The most current release, Stata 5.0, is available on the arizona UNIX servers. Stata is also installed on personal computers in a number of CIT and private clusters. There is also a version of Stata which is available for the Macintosh.
Stata can be used to enter and edit interactively data, and for both simple and complex statistical analysis. Commands can be executed one at a time at the Stata prompt, or groups of commands can be entered into do-files which can then be executed.
Stata 5.0 is available on the arizona UNIX servers. To invoke Stata 5.0 issue the following command at the UNIX prompt:
flagstaff.Princeton.EDU% stata
If you are working in a telnet session, this should be what you see: :
On PC's, you can invoke Stata by clicking on the Stata icon and your screen should look like this: :
We will refer to "Stata for PC" features using the symbol that appears at the upper-left corner of your screen, and these features will be presented in red.
________________________________________
Data Management
Data management is perhaps the most important part of your work in Stata. It is how you get your data into a format in which you can run your analyses properly. Major mistakes have been the result of poor data management rather than poor analysis.
Reading Data
There are three basic ways of reading data into Stata. "use" is for reading data that have been saved in Stata format, "insheet" is for spreadsheets saved as "CSV" files from a package such as Excel, "infile" is for data in what are called "flat" files, and "dictionaries" are for when you have data that has to be read in using special formats. .
Use
The Stata use command reads data that has been saved in Stata format:
use [filename]
where filename is the name of the Stata file. For example, suppose you have a Stata file named "myfile.dta." You can read this file with the following command:
use myfile
Note that the ".dta" file extension is automatically appended to Stata files. You do not have to include the file extension on the use command. Also note that the file should be stored in your working directory. You can also read files from other directories, e.g.:
use \home\project\myfile
On a PC, use the "File/Open..." menu for reading Stata formatted data. Stata will automatically clear your memory from any existing data.
Insheet
This is an example of a spreadsheet from Excel. It cannot be brought into Stata without some editing and saving it as a ".csv" file.
The insheet command is very useful in reading in data from a spreadsheet. There are some conventions required though:
1. The first line should have Stata variable names (eight characters or less, not starting with a "special character" or number) and the second line begins the data.
2. Missing numeric data should be coded as an empty cell, not a space, dot, or any other non-numeric data. Often, 0, 9, or 99 is used to code missing numeric data; this is fine as long as these are not also valid values for that variable.
3. Commas in numbers or text are particularly problematic because Stata thinks they are a delimiter and will not read the data properly. You must remove the commas from numeric values before saving the file.
4. When some spreadsheets create a .csv file it does not add commas to the end of a line if the cells at the end of that line are empty. This will confuse Stata which relies on the commas to tell it where the values are. You can avoid this problem by adding another column of 1’s (or any other character) to your spreadsheet. You can drop this variable once you have read it into Stata.
5. The file must be specifically saved as a "comma separated values" file in Excel. You can do this by going to "File", then "Save As…", then choosing "comma separated values." Simply giving it an extension of ".csv" will not work. When you close the spreadsheet and Excel asks if you want to save the changes, say "No." This is counter-intuitive, but the changes it’s asking you about are the changes it needs to make the spreadsheet a regular Excel spreadsheet again.
Once you have completed the changes and you are in Stata, give the command:
. insheet using filename.csv
to read in the file. If you get an error message about "wrong number of values," then you have a problem with not enough commas in the file, see #4 above.
Infile
"infile" may be useful if you have downloaded a data file from the web and it conforms to these specifications:
1. The file should NOT have variables names on the first line.
2. Character variables that have spaces in them, such as full names, must be enclosed in quotes.
3. Numbers can have commas and minus signs, but not dollar or percent signs.
4. Infile assumes that the variables have spaces between them and that there are no blanks spaces where it expects data (missing data needs to be represented by something).
The command to read a file with infile is:
infile var1 var2 var3 using mydata.raw
Essentially, insheet assumes that commas separate the variables, and infile assumes that blanks separate the variables.For example, suppose you have a raw data file named mydata.raw which contains the following fields:
1 55 4.5
2 23 3.2
3 34 3.4
4 52 7.1
5 41 2.9
You can use the following Stata command to read this file:
infile [varlist] using [filename]
where varlist is a list of variables and filename is the name of the file that contains the raw data. For example:
infile caseid age score using mydata.raw
where caseid, age and score are variable names and mydata.raw is the file that contains the data. It may be easier, though, to bring a file into Excel first, save it as a csv, then use insheet.
Moreover, you can use the infile command to read data files containing many variables or multiple records per observation, or more generally when you need to read data that have been saved with particular formats. In these cases, it is often much easier (and sometimes necessary) to read data using a Stata dictionary file, as described below.
Dictionary Files
Stata dictionary files can vary in their complexity but all of them basically specify which data file to read, where the variables are located in the file and what they are to be called. Optionally, labels can also be provided in the dictionary file. Once the dictionary file is written, it is accessed by issuing the Stata infile command:
infile using [dictionary_file]
where dictionary_file is the name of your Stata dictionary file.
In this example, the raw data are saved in a file named "latimes.raw" and the dictionary file is saved in a different file named "latimes.dct." To read the data, you would call the dictionary file:
infile using latimes.dct
Given below is a sample Stata dictionary file:
dictionary using latimes {
_column(13) wght %5f "Weight Value"
_column(22) groups %1f "Ethnic/racial grp"
_column(23) rdir %1f "Cntry in right dir.?"
_column(24) apbush %1f "Do You app. of Bush?"
_column(27) rvote %1f "Are you reg. to vote?"
_column(28) wvote %1f "Vote for/Lean toward?"
_column(29) whyvote %1f "Why this candidate?"
_newline
_column(48) location %1f "Area you live in?"
_column(49) views %1f "Views on political matters?"
_column(50) tparty %1f "Party views?"
_column(52) relign %1f "What religion?"
_column(55) faminc %1f "Family income"
_column(66) agerng %1f "Age range"
_column(69) educ %1f "Education"
_newline
}
The above dictionary starts with the words dictionary using , which defines the file as a Stata dictionary. The name of the file which contains the actual data appears immediately following the word using . In this example, the data file is named "latimes.raw" and the name latimes appears in the dictionary file. The data file extension is assumed to be .raw unless a different file extension is specifically provided.
The actual data dictionary is enclosed within curly brackets ({ }). On each line of the dictionary, the _column command is used to move the pointer to the location where the data item starts. After the column locator is the variable name. After the variable name is the format with which the data item is to be read (see the manual for possible formats). After the format, a quoted text string can be provided for a variable label. If there is more than one line of data for each record, a _newline command must be given to advance the pointer to the next line in the data file.
How to Save Stata Files
You can use the Stata save command to save a file in Stata format:
save [filename]
where filename is the name of your Stata file. For example:
save myfile
will save a Stata file named "myfile.dta" in your working directory. This file can be read in Stata with the use command. Note that the ".dta" file extension is automatically appended to Stata files. You do not have to include the file extension on the use or save commands.
If you already have a Stata file named "myfile.dta" and wish to save an updated version of the file under the same name, then use the Stata save command with the replace option, as in:
save [filename], replace
where filename is the name of the file you which to replace, e.g.:
save myfile, replace
To save an updated version of the active file, you can simply type:
save, replace
To save your file in some other directory, you can type:
save \home\project\myfile, replace
This command will destroy the previous version of your file so use the replace option only if you are certain that you will not need the older version of your file. There is no way to retrieve your original file once another file has written over it.
In Stata for PC, to save the file "myfile.dta" use the "File/Save As..." menu.
How to Increase Memory
Sometimes you may need to allocate additional memory for your Stata session, such as when you are working with a large file. If you recieve this message from Stata:
no room to add more observations
then you should increase the amount of memory available to your Stata session. Here's how.
1. Find out how large the file is. First, issue the clear command to remove the file from memory. Then issue the desc using filename command:
desc using mydata.dta
At the top of the information listed is the size of the file, in bytes. There are 1,000 bytes in a kilobyte, and 1,000 kilobytes in a megabyte, so if the size is 11,000 then the file is 11 kilobytes. For example:
Contains data
obs: 1,248
vars: 104
size: 11,576,000
-------------------------------------------------------------------------------
This shows that the file has 1248 observations, 104 variables, and is just slightly over 11.5 megabytes.
2. From the Stata dot prompt, issue the command set memory to increase the amount of memory. For example, the following command allocates 12 megabyes of memory to the current Stata session:
set memory 12m
Set the memory to a number slightly larger than the size of the file you are trying to read.
3. Now read your data file.
The memory and set memory commands are new features of Stata 5.0 for reporting, allocating and repartitioning memory. The memory and set memory commands in release 5.0 replace the following commands in release 4.0: the memsize command and the set commands for memsize, maxvar and maxobs.
• Use the Stata memory command to determine how much memory is currently being used and how much is available. For example, when the following command is issued at the Stata dot prompt:
memory
The following report will be displayed:
Total memory 2,097,144 bytes 100.00%
overhead (pointers) 0 0.00%
data 0 0.00%
------------
data + overhead 0 0.00%
programs, saved results, etc. 352 0.02%
------------
Total 352 0.02%
Free 2,096,792 99.98%
This reports indicates that more than 2M bytes of memory are available for the current session, the default allocation in UNIX.
• Use the set memory command to allocate additional memory, if needed. The general syntax is:
set memory [memsize]
where memsize is the amount of memory requested. For example, issue the following command to request 4 MB of memory for your current Stata session:
set memory 4m
How to Reduce Memory Requirements
You may be able to reduce your memory requirements by saving your data more efficiently. You can use the compress command to reduce the amount of memory your data consumes.
This command does not create a compressed version of your file in the way that compression utilities do, such as zip or pkzip. Rather, the Stata command compress changes the data types used to store your variables such that each variable is stored optimally.
In numerous examples, memory reqirements have been reduced by as much as 85% using the compress command.
1. Check the size of the original file using the UNIX ls command. This command can be issued from within a Stata session:
ls -l myfile.dta
where myfile.dta is the name of your Stata file. The -l (long) option displays a long listing of the file that includes its number of bytes. In Windows, you can ude the DOS dir command:
dir myfile.dta
2. Open (use) your Stata file. For example:
use myfile
(Recall that the .dta file extension is assumed on the use command.)
3. issue the compress command:
compress
Stata will consider changing the data type for each of your variables. A message will be displayed if the storage type is changed.
4. You need to save the new version of the file in order for the changes to remain in effect. Issue the Stata save command followed either by a new name for your Stata file or by the replace operand. For example
save myfile2
will save a new Stata file named myfile2.dta. You will have the original file, myfile.dta plus the new file.
Alternatively, the following command will overwrite the original file:
save, replace
This will save the data is in Stata's work area as a file named myfile.dta, replacing your original data file. You should be certain that you want to overwrite your original Stata file before using the replace option.
5. Now for the fun part. Check the size of the new file using the UNIX ls command. This command can be issued from within a Stata session:
ls -l myfile2.dta
where myfile2.dta is the name of your Stata file. You may see a significant decrease in the size of your file. You need to compress your Stata file only once.
Examining your Data
Describe
Once you have the data in Stata, you will want to make sure that all the variables are there and that they are in the format you need. You can do this with the "describe" command. Describe, which can be abbreviated as simply "d," will provide basic information about the file and the variables. You don’t have to call the data into Stata to be able to describe it, though. The command:
d using myfile :
will accomplish this. This can be very useful if the the file is too large to fit into memory. Often, your dataset will be larger than the default and you will need to increase the amount of memory Stata uses. To do this, look at the "size" line in the describe output; this is the size of the file in bytes. Since 1,000,000 bytes is a megabyte, files larger than 4,000,000 will not be loaded into Stata unless you increase the memory with the "set memory" command:
set mem 10m
This example increases the memory to 10 megabytes. Make sure that you give yourself some extra memory so you can create new variables and/or add observations.
List
It is often useful to just look at the data without doing any kind of analysis. The "list" command, abbreiviated as "l" will let you do this. Simply giving the "l" command will display all of the data on your screen, if you specify certain variables in the command, then only those variables will be printed:
l :
l var1 var2 var3 - displays just var1, var2 and var3 :
In Stata for PC you can have a direct look at your data as if you were working with a spredsheet (e.g. Excel), by using the "Stata Editor". You enter the editor by clicking on the "Editor" button. :You can edit your data by highlighting the cell you want ot change, entering the new value for that cell, and clicking on "return". You can edit the variable name (and the variable label) by double clicking on the relevant column : .
Creating Variables
Stata can store data as either numbers or characters. Stata will allow you to do most analyses only on numeric data. Sometimes, when you use insheet or infile, a numeric variable gets read in as a character, or string variable. This may be due to a ‘space’ character or a ‘.’ in one of the cells. Since Stata allows you to do analyses only on numeric variables, you will need to convert them to numeric data.
There are different types of numeric variables - float, binary, double, long and int - the differences among them are simply how much space they take up in the file. In most cases, you will not need to concern yourself with these differences.
Encode
Use encode when the original variable is, indeed, a character variable (such as gender being coded as "m" and "f’") and you need numbers instead. The encode command does not produce dummy variables, it just assigns numbers to each group defined by the character variable. In this example, gender was the original character variable and sex is the new numeric variable:
encode gender, gen(sex) :
Generate
The "gen … real" command will create a numeric variable based on the original string variable. Use this if the original was defined as a character variable "by mistake." The following command will create a new, numeric variable var1n by converting the string variable, var1:
gen var1n=real(var1)
Often, you will need to create new variables based on the ones you have already. The two most common ways of creating new variables is by using "generate" and "egen." Here are some examples of gen:
gen total= var1n + var2 + var3
gen cumtot=sum(total)
gen id=_n
In the first example, we generate a new variable called "total" which is simply the addition of var1n , var2, and var3. In the next example, we create a cumulative sum of total (the value of cumtot for this observation is the sum of total for all previous observations). Finally, we create an "id" variable by simply setting it equal to "_n", which is Stata’s way of numbering the observations in a dataset. This id variable can be very useful when you need to sort your data in different ways (see sorting below). The variable "_n" can be used at any time, but you must remember that it refers to the number of the observation in the current order, not the original order. Here are some more examples of gen and _n:
gen total2=42.3 if _n==125
gen lagtot=total[_n-1]
In the first example we simply set total equal to a specific value for the first through the 125th observation, namely, observation number 125. In the second, we create what is called a "lagged" variable, lagtot, which is the value of total from the previous observation designated by the "[_n-1]." You can use any valid arithmetic expression inside the brackets. Here are two last examples:
gen price = total + lagtot if company=="IBM"
replace price = total + cumtot if company~="IBM"
Here, we generate a new variable, "price", which is the sum of total and lagtot, but only if the value of company is IBM. Note two things here: first is the double "=" after "if", and second, the use of the quotes around IBM. When you refer to string variables in equality expressions such as this, you must put the quotes around the value so Stata will not confuse it with a variable name. Had we left the quotes out, Stata would have thought we wanted to generate price for observations where the variable company equals the variable "IBM." In the second example we use "replace" instead of "gen." This reason for this is that in the first example, Stata creates values only for those observation fulfilling the "if" condition. All other observations get "missing" values. We want, then, to replace those missing values with something. The "~" in Stata means "not," and suffices as the extra "=."
Dummy variables
Sometimes we need to generate a "dummy" variable, or variables. Stata makes this very easy:
tab size, gen(dummy) :
Here, Stata will create a dummy variable for each value found in size. So, dummy1 = 1 if it is a big company, 0 otherwise; dummy2 = 1 if it is a medium company, 0 otherwise; dummy3 = 1 if it is a small company, 0 otherwise. (See a list and desc)
If you want a dummy variable to indicate only a particular size category:
gen d_med = (size==2)
Here Stata will create a dummy variable such that: d_med = 1 if it is a medium company, 0 otherwise. In general, using this command will make Stata create a dummy variable equal to 1 for each observation where the expression in brackets is true, and equal to zero otherwise.
Character Variables
Stata assumes that the variable you are creating is numeric unless you tell it otherwise. Sometimes, though, you will want to create a variable whose values are strings. Here’s how:
gen str5 name="John"
gen str10 team="Atlanta" in 1/5
As above, the values for string variables must be enclosed in quotes. You must also specify the largest number of characters, including spaces, that the variable will have in the "str#". In other words, if "John" is the longest value for the variable name, then you must specify name to have at least 4 characters. It’s fine if you specify more characters than the values will actually take, but if you specify less, you will either get an error message, or some of the characters will be cut off.
The "in" option simply tells Stata to apply the "gen" command to observations 1 to 5, inclusive, in the order they appear in the dataset. You can specify any consecutive range. The "in" option can also be used with numeric variables, as well as any other commands.
gen str3 yn="yes" if team=="Atlanta" & name=="john"
replace yn="no" if team=="New York" | team=="Boston"
Here we extend the use of the "if" option to include two expressions. As you may have guessed, the "&" means "and." The "|" means "or." One thing you must be wary of is that the values of character variables are case-sensitive, so "John" is not the same as "john."
gen day=substr(code,1,2)
The substr function can be helpful; it specifies certain parts of a character variable to be used. The first argument specifies the variable to be picked apart (code), the second argument is the starting character, and the last argument is the number of characters to be pulled out. In the first example, starting with the first character in the variable "code," two characters will be extracted.
Date Variables
Stata stores dates as the number of elapsed days since January 1, 1960. There are two ways to create elapsed dates, depending on how your original variables are stored. If you have separate variables for month, day and year then use the mdy() function to create an elapsed date variable. If your original dataset already contains a date variable, then use the date() function. Once you have created an elapsed date variable, you will probably want to format it, as described below.
• Use the mdy() function to create an elapsed date variable when your original data contains separate variables for month, day and year. The month, day and year variables must be numeric. For example, suppose you are working with these data.
month day year
7 11 1948
1 21 1952
11 2 1994
8 12 1993
Use the following Stata command to generate a new variable named mydate:
gen mydate = mdy(month,day,year)
where mydate is an elapsed date varible, mdy() is a Stata function, and month, day, and year are the names of the variables that contain data for month, day and year, respectively. The mdy() function can put three variables together to create a new date variable. The variables do not have to be called month, day, and year, but they do have to follow that order. One more caveat is that the year must be stored as a four-digit number, otherwise Stata will not know what century you want. You may need to do some more programming to add "1900" to the value.
• Use the date() function to create an elapsed date variable when your original dataset contains a date variable. For example, suppose you have a string variable, oldate, that represents the date of some event.
olddate
7-11-92
3-24-93
9-16-94
12-18-1996
Use the following Stata command to generate a new variable named newdate:
gen newdate = date(oldate,"mdy")
where newdate is an elapsed date varible; date() is a Stata function; olddate is the name of the variable that contains a date; and "mdy" is the format for reading the data (i.e., month, day and year).
Other formats are available for the date() function including "dmy" and "ymd". The format must match the order in which the values appear in the original date variable. You can use only one format to read a given date variable since the month, day and year components must appear in the same order for every occurance of the original date variable. However, you can represent the month, day and year components in a number of ways, for example:
somedate
7/11/92
Sept 24, 1993
9.16.94
December 18, 1996
The date() function can be used to read these data and create an elapsed date variable named newdate2:
gen newdate2 = date(somedate,"mdy")
• Use the format command to display elapsed dates as calendar dates. In the example given above, the elapsed date variable, mydate, has the following values, which represent the number of days before or after January 1, 1960.
month day year mydate
7 11 1948 -4191
1 21 1952 -2902
8 12 1993 12277
11 2 1994 12724
You can use the format command to display elapsed dates in a more customary way. For example:
format mydate %d
where mydate is an elapsed date variable and %d is the format which will be used to display values for that variable.
month day year mydate
7 11 1948 11jul48
1 21 1952 21jan52
8 12 1993 12aug93
11 2 1994 02nov94
Other formats are available to control the display of elapsed dates (see the manuals).
Here are some other useful functions for working with dates:
gen day = day(date)
gen weekday = dow(date)
gen month = month(date)
gen year = year(date)
All of these examples assume that the date variable is an elapsed date. The day() function returns the number of the day, ie, 23. The dow() returns a number from 0 to 7: 0 if it is a Sunday and 6 if it is a Saturday. month() returns a number from 1 to 12. year() returns the year as a four-digit number. :
Extended Generate (egen)
"egen", or "extended generate" is useful when you need a new variable that is the mean, median, etc. of another variable, for all observations or for groups of observations. Egen is also useful when you need to simply number groups of observations based on some classification variables. Here are some examples:
egen sumvar1 = sum(var1)
egen meanvar1= mean(var1), by(var3)
egen count = count(id), by(company)
egen group = group(month year)
In the first example, we simply create a variable whose value for each observation is equal to the sum of var1 for all observations (all observations will have the same value). The second example shows how to create a variable that is the mean of another variable for each group designated by var3 (all observations within a group will have the same value). The third example simply counts the number of non-missing id values in each company (this basically counts the number of observations you have for each company). The last example simply assigns a number to each of the groups created by the combination of month and year. :
There are many functions that can be used with "extended generate". To have a brief summary, type:
help egen
Variable and Value labels
Now that we’ve created all these new variables, we’ll want some way of keeping track of what each one is and what their values mean. We can do this by creating variable labels and value labels. These labels are not necessary in Stata, they just make the output easier to read. Variable labels correspond to the variable names, whereas value labels correspond to the different values a variable may have. Here’s how to create them:
label variable var1 "The first variable"
This assigns a label to the variable, var1, so whenever the variable name var1 is displayed on the screen, the description "The first variable" will be displayed as well.
In Stata for PC, variable names and variable labels will automatically appear in the "Variables" window, and you can enter a variable name in the "Stata Command" window simply by clicking on it in the "Variables" window. :
Variable labels must be 31 characters or less. Often, though, it’s more helpful to label the values a variable can take so you don’t have to memorize them or keep referring to a codebook.
label define grp 1 "Male" 2 "Female" 3 "N/A"
label values gender grp
Value labels correspond to the actual number or letters in the data. They are used so that the printout will show "Male," "Female," and "N/A" instead of "1," "2," and "3". First you have to "define" the label, then associate that label with a variable. Each label can be associated with more than one variable, so if there are several "yes/no/maybe" questions, you just have to define one value label and use it for all the questions. You may associate only one value label with a particular variable, however. :
You can check the list and content of all value labels by typing:
label list
Keep, Drop, and Rename
Well, now we’ve created many new variables and converted some old ones. Since we no longer need all of these variables, we’ll want to eliminate some of the ones we don’t really need and perhaps rename some of the ones we keep. We can either keep the variable we are interested in:
keep code date var4
or we can drop those we don't need, and rename some of those we keep:
drop var1 var2 var3
rename code date
Here we drop the variables, var1, var2 and var3. Be careful! Once they are dropped, they can’t be picked up again unless you clear the data and use it again. You can drop as many variables as you want in one command. The same is true for "keep". The second example renames the variable "code" to "date." You can rename only one variable at a time. You can also drop (or keep) observations:
drop if size==1
or
keep if size~=1
will drop all observations whose size is equal to 1. :
________________________________________
Log and Do Files
Log files
Log and do files are very useful. Logs keep a record of what commands you have issued and their results during your Stata session. Do files are good for long series of commands that may need to be "tweaked" to work properly. They are also necessary to replicate things that you have done on new or modified datasets.
You can create a log file with the command:
log using filename
Where filename is any name you wish to give the file. You will find it helpful to use names that will help you to remember what you did during that session. Stata will automatically append an extension of ".log" to the filename. By default, everything displayed on the screen will be recorded in the log file. You can give the file any name you like, but you should use names that will help you remember what analyses you did.
If you work for a long time, this file can get to be very large and have quite a bit of unnecessary output. So, if you are trying different things to "see what happens," you may want to start and stop the logging several times in one session. You can stop the logging with the command:
log close
If you turn the logging on and off during a session, you may need to use the append and/or replace options:
log using filename, append
log using filename, replace
The append option simply adds more information to the file, whereas the replace option erases anything that was already in the file. Be careful! It’s often helpful to put comments in the log file to help you remember why you did something. You can do this by simply preceding anything you type with a "*".
But, what happens if you forget to start a log file? There is a way to save what you have done by using the log file and the #review command. The #review command will list the last however many commands you have issued. This command can be used at any time and you can specify any number of commands to retrieve. After some minor editing and changing the extension to ".do", you can run it as a do file. Here’s an example:
log using iforgot - start a log file
#review 20 - list the last 20 commands
log close - close the log file
The file, iforgot.log, will have the last twenty commands you issued in it. You can then run this file as a do file with a new log started.
In Stata for PC you open a .log file by clicking on the "Log..." button. : If you want to open an existing .log file, you have to choose whether to overwrite it or to append the output to its existing content :. At this point a "Stata Log" window will pop up, and the contents of the .log file will be visualized there :. Once the "Stata Log" window is open, every command you entere "Stata Command" window will have its output saved in the log file. By clicking on the "Log..." buttom you can choose either to keep the log file open and suspend saving the output in it, or close the log file :. You can also print your log file by opening it and using the "File/Print Log" menu.
Do Files
A "do" file is a set of commands just as you would type them in one-by-one during a regular Stata session. Any command you use in Stata can be part of a do file. Do files are very useful, particularly when you have many commands to issue repeatedly, or to reproduce results with minor or no changes.
You can use any editor you wish to create your do files. If you use pine for email, then you already know how to use pico. The only thing you need to remember is that if a command is longer than one line, then you need to use the #delimiter command. In Stata, hitting the return key tells Stata to execute the command. In a .do file, the return key is at the end of every line, so you need a way of telling Stata that the command is longer than one line. Here is an example of a short do file we’ll call mydofile.do that you create in a text editor:
log using mydofile - start a log file
#delimiter ; - set the delimiter to ";"
use mydata; - open the data file
des; - describe the file
collapse (mean) var1 var2 var3 (meadian) medvar=var1, by(var4);
- collapse the data
save mydata2; - save the collapsed data
#delimiter cr - set the delimiter back to the "enter" key
clear - clear the data from memory
log close - stop logging
To run this file in Stata, you would simply issue the command:
do mydofile
You don’t even need to exit Stata to cretae or edit the do file. The command:
!pico myfile.do
will allow you to enter the text editor, pico, without exiting Stata. Once you save the file and exit from pico you are automatically retuned to Stata. You can use the ! character to tell Stata that you want this command issued to UNIX (or to DOS if you are working with Stata for PC in Windows), rather than Stata itself. Any command you would normally issue to UNIX (or to DOS)can be issued from within Stata this way. The "ls" (list files) and "cd" (change directory) commands do not have to be prefixed by the !.
Two other useful things are the Ctrl-r and Ctrl-c keys. You can use Ctrl-r to recall on the command line the commands you have issued. Sometimes the output from a command is very long and you don’t need to look at all of it; you can use Ctrl-c to stop the output.
In Stata for PC, run a .do file by using the "File/Do..." menu : . You can edit it with any normal text editor ( e.g. )
________________________________________
File Management
Once you have all of your variables in the format you want, you’ll need to get the entire file in a format that will make it easier to use. You can do this by sorting, appending, merging and collapsing.
Sort
"sort" puts the observations in a data set in a specific order. Some procedures require the file to be sorted before it can work. You can sort a file based on more than one variable. One thing you must be careful of, though, is that Stata will randomize the order of the observations within the variables used to sort. This is why creating the id variable mentioned earlier is important. With the id variable, you will always be able to go back to the original order and start over.
sort month :
sort month year
In Stata for PC you can sort data in the Stata Editor, by clicking on the variable you want to sort by, and then clicking the "Sort" button.:
Append
Sometimes, you have more than one file of data which you need to analyze. One case may be that you have two files with the same variables but different observations. The other case is when you have two files with the same observations, but different variables.
Use append when you simply want to add more observations, in other words you already have data for 1990, and now you want to include new observations from 1991 for the same variables. For example:
use class90
append using class91 :
will add the observations from the file, class91 (what Stata refers to as the "using" dataset), to the end of class90 (what Stata refers to as the "master" dataset). Any variables with different names in the two files will have missing values for the observations from the other dataset.
Merge
If you have two files that have the same observations, but different variables, then you’ll want to "merge" them so you can use all of the variables at once. When you merge datasets, you are adding new variables to existing observations rather than adding observations to existing variables. There are two basic kinds of merges, a one-to-one and a match.
A one-to-one merge simply takes the two files and puts them side-by-side, regardless of whether the observations in each dataset are in the same order. A simple one-to-one merge isn’t very common, and isn’t recommended, even if you think all the observations match. Sometimes there may be a "glitch" that can throw off the order of the data. It’s just as easy to do a match-merge, and you’ll be able to check that all the observations matched correctly.
There are some "rules" when doing a match-merge. First, each dataset must have a "key" variable by which the observations can be matched - social security number is a good example. Second, both datasets must be sorted by this key variable. You can use more than one key variable if you want, such as month and year. These key variables must be of the same type (string or numeric) in both datasets. By default, if a variable is present in both datasets, then the values in the master dataset will remain unchanged. If the variables in the using dataset have the same names as ones in the master dataset, but represent additional information, then you will need to rename them in one of the datasets before you can merge them.
use class90
sort date
merge date using class91 :
In this example we simply match observations using the variable "date" as the key.
Sometimes, you will want to update the data you have, namely, fill in missing values. Using the update option will accomplish this:
merge date using class91, update
Values in the master dataset will be changed only if they are missing.
In some instances, you may want to replace non-missing values in the master dataset. Use the replace option in addition to the update option to do this:
merge date using class91, update replace
"replace" will not work by itself, it must always be used with update. Stata will not, under any circumstances, change a non-missing value in the master dataset with a missing value from the using dataset. You must use the "replace" command discussed earlier to do this.
Stata will automatically create a variable called _merge which will indicate the results of the merge. Always check this variable to make sure that you got what you wanted. Here are what the possible values of _merge mean:
1 = Observations from the master dataset that did not match observations from the using dataset.
2 = Observations from the using dataset that did not match observations from the master dataset.
3 = Observations from both datasets that did match.
4 = Observations from both datasets that did match, missing values in the master dataset were updated.
5 = Observations from both datasets that matched, values in the master dataset disagree with those in the using dataset.
Usually, you will want all the observations to have a value of 3. Values of 4 occur only when you use the "update" option. Values of 5 occur only when you use the "update" and "replace" options. If you need to merge another dataset, you have to re-name or drop _merge first, other wise you will get a "_merge already defined" error. :
Collapse
"collapse" is used when you want to create a dataset containing the means sums, etc., of the various groups in the data. One example might be when you have one dataset of monthly data and another of yearly data and you need to analyze both sets of information together.
collapse (mean) var1 var2, by(month)
This will create a dataset of the means of var1 and var2 by month. If you have twelve months in your original dataset, then you will have twelve observations in the collapsed dataset. You must be careful, though, because Stata will compute the statistics on a variable-by-variable basis. If one variable has more missing observations than another, the means (and any other statistics you request) will be based on a different number of observations. This is not always an acceptable practice. To avoid this, you must use the "cw" options for "casewise deletion." This means that Stata will drop any observation that does not have data for both var1 AND var2, thereby ensuring that all the statistics will be based on the same number of observations.
collapse (mean) var1 var2, by(month) cw :
________________________________________
Basic Commands
Now that you have your data in a format you want, check it before doing any analyses. This can save you quite a bit of frustration later on.
Summarize
"sum", short for summarize, will give you the means, sd’s, etc. of the variables listed. If you don’t list any variables, it will give you the information for all numeric variables. If a variable you thought was numeric shows up as having 0 observations and a mean of 0, then, most likely, Stata still thinks it’s a character variable.
summ :
summ var1 var2
The "detail" option gives you additional information about the distribution of the variable.
summ var1 , detail :
Inspect
"inspect" is another easy way to eyeball the distribution of a variable.
inspect var1 :
Tabulate
"tab", short for tabulate, will produce frequency tables. By specifying two variables, you will get a crosstab. There are other options to get the row, column and cell percentages as well as chi-square and other statistics; check the manual.
tab var1
tab var1 var2
tab var1 var2, row col chisq :
"By" group processing
Sometimes you’ll want to run a command or analysis on different groups of observations. The "by variable:" subcommand is the same thing as running the command with separate "if" statements for each group. You must sort the data before you can use the "by:"
sort sex
by sex: summ var2
by sex: gen lag2=var2[_n-1] :
Sometimes, this can produce more output on the screen than you would care to look at. To suppress output to the screen, but not to the log file, use "quietly."
quietly by var1: gen lag2=var2[_n-1]
________________________________________
Fancy commands
Macro variables
Sometimes you need to use many variables the same way many times. One example would be if you want to run regressions on different dependent variables using the same set of independent variables. This can mean a lot of typing. One way around this is to create a macro variable. A macro variable is simply a variable that has as its value a particular string. This string can be anything you specify: a list of variables, a particular command, or whatever. Whenever Stata comes across the macro variable, it will interpret it to mean whatever string you set the variable to.
local macvar var1 var2 lagtot lag2
reg depvar `macvar’
In this example, when Stata "sees" the variable macvar in the regression command, it replaces `macvar’ with the string "att att2 itt itt2 date." Pay careful attention to the different type of quotation marks: the first quote in `macvar’ is the opening left quote usually found under the ~. The second quote is the closing right, or single quote usually found under the ".
For
Other times, you will want to perform the same command on several variables. Again, this can mean a lot of typing. The "for:" command can also be very useful in these situations. It can use several types of variable lists, and with a little ingenuity, can be very powerful. Here are a few examples:
for var1-var25: replace @=. if @==99
for var*: replace @=. if @=99
for 1-3, ltype(numeric): gen q@=0
for a b c, ltype(any): gen str2 @="x"
The @ in the for command represents the variable names in the list, one by one. In the first two examples, Stata would interpret the command as though you had typed replace var1=. if var1=99, replace var2=. If var2=99, and so on.
In any command, you can specify two kinds of lists:
1. var1-var25 or date-q25 indicates all variables between var1 and var25, inclusive, even if they do not begin with "var" as found when you run the describe command.
2. var* indicates all variables starting with "var."
in the third example, the command will create three new variables called q1, q2, and q3 and sets them all equal to zero. The last example creates three new variables called a, b, and c, and sets them all equal to "x."
________________________________________
Regression
Linear Regression Model
In Stata you can estimate Linear Regression Models using the "regress" command (shortcut reg). Here we suppose that you are familiar with linear regression models, and more in general with multiple regression analysis, and thus we don't provide any introduction to this subject.
Now, suppose your dependent variable is depvar and your independent variables are listesd in varlist. Then the command to run the linear regression is:
reg depvar varlist
The first variable after the regress command is always the dependent variable (or left-hand-side variable), and the following list gives the relevant independent variables (or right-hand-side variiables).
Suppose, for example, that you have a data set with individual observations on earnings, and you are interested in regressing the natural logarithm of average weekly earnings (lnw) on a sex dummy variable (sex: 0=male, 1=female), on a race dummy variable (race: 0=white, 1=non-white), on years of schooling (school), on age (age) and on a constant. Then you should type
reg lnw sex race school age :
Notice that Stata automatically adds the "constant" (_cons) to the list of independent variables. If you want to exclude it, you can use the option nocons:
reg lnw sex race school age, nocons
As for many other commands in Stata, the regress command can be also applied to a subset of your data set by specifying an if (or in) statement. Suppose that you want to run a regression only for the male population:
reg lnw race school age if sex==0 :
Several options are available with the regress command (see help regress). Among the most useful is the robust option, that allows you to run the regression using a "robust" estimator for the variance (specifically, the Huber/White/sandwich estimator). This option is mainly used when the residuals are thought to be heteroschedastic:
reg lnw sex race school age, robust
Another option that may be useful is the one that allows you to run Instrumental Variable regressions (or 2SLS). Suppose you have a dependent variable y1 and a list of independent variables x1 x2 x3; and let's suppose that you want to instrument x3 with z1, x1 and x2. Then you can simply run the (2SLS) regression by typing:
reg y1 x1 x2 x3 (z1 x1 x2)
and you will get the results of the second stage in the 2SLS procedure.
Logit, Probit
When you dependent variable is dichotomous (zero-one), you can run a Linear Probability Model (using the regress command) or you might be interested in running a Logit or a Probit regression. Here we assume that you are familiar with the Logit and Probit models, as well as with maximum likelihood estimation. As a quick reminder, the only difference between the Linear Probability Model (LPM), the Logit model and the Probit model is the assumption that you make about the probability distribution of your dependent variable. Let y be your dependent variable (dummy: zero-one), and let x and z be your independent variables. Then the general specification in all these three models is
Prob (y = 1) = F (bx + cz), Prob (y = 0) = 1 - F (bx + cz)
where b and c are the coefficeints of x and z, and F(.) is some function. The LPM assumes that F(a) = a. The Logit model assumes that F(.) is the logistic distribution function, while the Probit model assumes that F(.) to be the normal distribution function.
To run a Logit regression in Stata, you simply type:
logit depvar varlist
where depvar is your dependent (dichotomous) variable, and varlist is the list of independent variables that enter the model. To run a Probit model, the command is similar:
probit depvar varlist
Suppose that you are trying to estimate the effectiveness of a new trug being test for cancer patients. Your dependent variable is died (1 if patient died, 0 otherwise), and your independent variables are age (age of patient at start of experiment) and placebo (1 if drug is placebo). Then your logit and probit estimates will be: : .
Stata reports the maximum likelihood estimates for the original coefficients in both the probit and logit commands. You might be interested in other "coefficients". The dprobit command estimates a Probit model, but reports the change in the probability for an infinitesimal change in each independent variable instead of the coefficient for each independent variable. The logistic command estimates a Logit model but reports the odds ratios for each independent variable instead of the coefficients. After both these commands, you can obtain the original coefficients by simply typing (respectively) probit or logit :.
The logistic command also offers several "post-estimation" commands that can used to evaluate the goodness of fit of the model. These commands are entered after the estimation is performed, and are always referred to the "last" logistic estimation perfomed. See help logistic for a list.
Finally, all the above commands (as in regress) can be applied only to a subset of the observations, can be run without constant, and can generate estimates using robust standard errors.
Prediction
After all estimation commands (i.e. reg, logistic, logit, dprobit, probit) several "predicted" values can be computed. The most important are the "predicted values for the dependent variable" and the "predicted residuals". In order to obtain the predicted values for the dependent variable for any model you have to generate a new variable (call it yhat) that associate to each observation its predicted value based on the coefficients obtained from the estimation and based on the particular values of the independent variables for that observation. Suppose that you run a linear regression and you want to estimate the predicted (ln) average weekly wage for each individual: then you type
reg lnw sex race school age
predict yhat
then a new variable called yhat will be generated by Stata (if such a variable already exists you will get an error message, and you will have to choose another name for the predicted values) : . In the same way you can build a new variable (call it res) with the predicted "residual" for each observation:
reg lnw sex race school age
predict res, residual
Notice that after the logistic command, several other "predictions" are possible using the lpredict command (see help lpredict).
Testing Linear Hypothesis
Here we assume that you are familiar with the theory of hypothesis testing. After each estimation, for each independent variable Stata automatically provides a t-test (for linear regressions) and a z-test (for Logit or Probit models) on the null hypothesis that the "true" coefficient is equal to zero. However you can perform several other tests on linear hypothesis about the coefficients. Suppose you run a linear regression of (log) average weekly earnings on sex, race, schooling and age. You could test the hypothesis that the coefficient on race is equal to -.05 (i.e. earnings for non-whites are 5% lower than for whites):
reg lnw sex race school age
test race=-.05 :
You could also be interested in testing hypotheses on several variables. For example you could be interested in testing the null hypothesis that the coefficients on sex and race are jointly equal to zero, or that they sum up to one:
reg lnw sex race school age
test sex race - tests the null that they are jointly equal to zero
test sex + race == 1 :
If you are interested in non-linear hypothesis testing, you can look at help testnl .
This page is designed to give you a quick look at some basic tasks you may need to do in Stata. What it does cover is:
1. Overview of Stata.
2. Creating and using "log" and "do" files.
3. Data management.
4. File management.
5. Basic data description.
6. Basics in regression analysis.
There are some other resources you may find useful:
1. Stata manuals
2. Stata Technical Bulletin
3. The DSS web site, www.princeton.edu/~data/stata (the major source for this overview)
4. Stata’s web site, www.stata.com
5. Stata’s on-line tutorial (the command, ‘help tutorial’ will show you what to do).
You can click on the to see what the output from that particular command or commands look like, or what a particular window should look like in the PC version of Stata; use the "back" button on your browser to return to this page.
________________________________________
Overview
Stata is an interactive data analysis program which runs on a variety of platforms. The most current release, Stata 5.0, is available on the arizona UNIX servers. Stata is also installed on personal computers in a number of CIT and private clusters. There is also a version of Stata which is available for the Macintosh.
Stata can be used to enter and edit interactively data, and for both simple and complex statistical analysis. Commands can be executed one at a time at the Stata prompt, or groups of commands can be entered into do-files which can then be executed.
Stata 5.0 is available on the arizona UNIX servers. To invoke Stata 5.0 issue the following command at the UNIX prompt:
flagstaff.Princeton.EDU% stata
If you are working in a telnet session, this should be what you see: :
On PC's, you can invoke Stata by clicking on the Stata icon and your screen should look like this: :
We will refer to "Stata for PC" features using the symbol that appears at the upper-left corner of your screen, and these features will be presented in red.
________________________________________
Data Management
Data management is perhaps the most important part of your work in Stata. It is how you get your data into a format in which you can run your analyses properly. Major mistakes have been the result of poor data management rather than poor analysis.
Reading Data
There are three basic ways of reading data into Stata. "use" is for reading data that have been saved in Stata format, "insheet" is for spreadsheets saved as "CSV" files from a package such as Excel, "infile" is for data in what are called "flat" files, and "dictionaries" are for when you have data that has to be read in using special formats. .
Use
The Stata use command reads data that has been saved in Stata format:
use [filename]
where filename is the name of the Stata file. For example, suppose you have a Stata file named "myfile.dta." You can read this file with the following command:
use myfile
Note that the ".dta" file extension is automatically appended to Stata files. You do not have to include the file extension on the use command. Also note that the file should be stored in your working directory. You can also read files from other directories, e.g.:
use \home\project\myfile
On a PC, use the "File/Open..." menu for reading Stata formatted data. Stata will automatically clear your memory from any existing data.
Insheet
This is an example of a spreadsheet from Excel. It cannot be brought into Stata without some editing and saving it as a ".csv" file.
The insheet command is very useful in reading in data from a spreadsheet. There are some conventions required though:
1. The first line should have Stata variable names (eight characters or less, not starting with a "special character" or number) and the second line begins the data.
2. Missing numeric data should be coded as an empty cell, not a space, dot, or any other non-numeric data. Often, 0, 9, or 99 is used to code missing numeric data; this is fine as long as these are not also valid values for that variable.
3. Commas in numbers or text are particularly problematic because Stata thinks they are a delimiter and will not read the data properly. You must remove the commas from numeric values before saving the file.
4. When some spreadsheets create a .csv file it does not add commas to the end of a line if the cells at the end of that line are empty. This will confuse Stata which relies on the commas to tell it where the values are. You can avoid this problem by adding another column of 1’s (or any other character) to your spreadsheet. You can drop this variable once you have read it into Stata.
5. The file must be specifically saved as a "comma separated values" file in Excel. You can do this by going to "File", then "Save As…", then choosing "comma separated values." Simply giving it an extension of ".csv" will not work. When you close the spreadsheet and Excel asks if you want to save the changes, say "No." This is counter-intuitive, but the changes it’s asking you about are the changes it needs to make the spreadsheet a regular Excel spreadsheet again.
Once you have completed the changes and you are in Stata, give the command:
. insheet using filename.csv
to read in the file. If you get an error message about "wrong number of values," then you have a problem with not enough commas in the file, see #4 above.
Infile
"infile" may be useful if you have downloaded a data file from the web and it conforms to these specifications:
1. The file should NOT have variables names on the first line.
2. Character variables that have spaces in them, such as full names, must be enclosed in quotes.
3. Numbers can have commas and minus signs, but not dollar or percent signs.
4. Infile assumes that the variables have spaces between them and that there are no blanks spaces where it expects data (missing data needs to be represented by something).
The command to read a file with infile is:
infile var1 var2 var3 using mydata.raw
Essentially, insheet assumes that commas separate the variables, and infile assumes that blanks separate the variables.For example, suppose you have a raw data file named mydata.raw which contains the following fields:
1 55 4.5
2 23 3.2
3 34 3.4
4 52 7.1
5 41 2.9
You can use the following Stata command to read this file:
infile [varlist] using [filename]
where varlist is a list of variables and filename is the name of the file that contains the raw data. For example:
infile caseid age score using mydata.raw
where caseid, age and score are variable names and mydata.raw is the file that contains the data. It may be easier, though, to bring a file into Excel first, save it as a csv, then use insheet.
Moreover, you can use the infile command to read data files containing many variables or multiple records per observation, or more generally when you need to read data that have been saved with particular formats. In these cases, it is often much easier (and sometimes necessary) to read data using a Stata dictionary file, as described below.
Dictionary Files
Stata dictionary files can vary in their complexity but all of them basically specify which data file to read, where the variables are located in the file and what they are to be called. Optionally, labels can also be provided in the dictionary file. Once the dictionary file is written, it is accessed by issuing the Stata infile command:
infile using [dictionary_file]
where dictionary_file is the name of your Stata dictionary file.
In this example, the raw data are saved in a file named "latimes.raw" and the dictionary file is saved in a different file named "latimes.dct." To read the data, you would call the dictionary file:
infile using latimes.dct
Given below is a sample Stata dictionary file:
dictionary using latimes {
_column(13) wght %5f "Weight Value"
_column(22) groups %1f "Ethnic/racial grp"
_column(23) rdir %1f "Cntry in right dir.?"
_column(24) apbush %1f "Do You app. of Bush?"
_column(27) rvote %1f "Are you reg. to vote?"
_column(28) wvote %1f "Vote for/Lean toward?"
_column(29) whyvote %1f "Why this candidate?"
_newline
_column(48) location %1f "Area you live in?"
_column(49) views %1f "Views on political matters?"
_column(50) tparty %1f "Party views?"
_column(52) relign %1f "What religion?"
_column(55) faminc %1f "Family income"
_column(66) agerng %1f "Age range"
_column(69) educ %1f "Education"
_newline
}
The above dictionary starts with the words dictionary using , which defines the file as a Stata dictionary. The name of the file which contains the actual data appears immediately following the word using . In this example, the data file is named "latimes.raw" and the name latimes appears in the dictionary file. The data file extension is assumed to be .raw unless a different file extension is specifically provided.
The actual data dictionary is enclosed within curly brackets ({ }). On each line of the dictionary, the _column command is used to move the pointer to the location where the data item starts. After the column locator is the variable name. After the variable name is the format with which the data item is to be read (see the manual for possible formats). After the format, a quoted text string can be provided for a variable label. If there is more than one line of data for each record, a _newline command must be given to advance the pointer to the next line in the data file.
How to Save Stata Files
You can use the Stata save command to save a file in Stata format:
save [filename]
where filename is the name of your Stata file. For example:
save myfile
will save a Stata file named "myfile.dta" in your working directory. This file can be read in Stata with the use command. Note that the ".dta" file extension is automatically appended to Stata files. You do not have to include the file extension on the use or save commands.
If you already have a Stata file named "myfile.dta" and wish to save an updated version of the file under the same name, then use the Stata save command with the replace option, as in:
save [filename], replace
where filename is the name of the file you which to replace, e.g.:
save myfile, replace
To save an updated version of the active file, you can simply type:
save, replace
To save your file in some other directory, you can type:
save \home\project\myfile, replace
This command will destroy the previous version of your file so use the replace option only if you are certain that you will not need the older version of your file. There is no way to retrieve your original file once another file has written over it.
In Stata for PC, to save the file "myfile.dta" use the "File/Save As..." menu.
How to Increase Memory
Sometimes you may need to allocate additional memory for your Stata session, such as when you are working with a large file. If you recieve this message from Stata:
no room to add more observations
then you should increase the amount of memory available to your Stata session. Here's how.
1. Find out how large the file is. First, issue the clear command to remove the file from memory. Then issue the desc using filename command:
desc using mydata.dta
At the top of the information listed is the size of the file, in bytes. There are 1,000 bytes in a kilobyte, and 1,000 kilobytes in a megabyte, so if the size is 11,000 then the file is 11 kilobytes. For example:
Contains data
obs: 1,248
vars: 104
size: 11,576,000
-------------------------------------------------------------------------------
This shows that the file has 1248 observations, 104 variables, and is just slightly over 11.5 megabytes.
2. From the Stata dot prompt, issue the command set memory to increase the amount of memory. For example, the following command allocates 12 megabyes of memory to the current Stata session:
set memory 12m
Set the memory to a number slightly larger than the size of the file you are trying to read.
3. Now read your data file.
The memory and set memory commands are new features of Stata 5.0 for reporting, allocating and repartitioning memory. The memory and set memory commands in release 5.0 replace the following commands in release 4.0: the memsize command and the set commands for memsize, maxvar and maxobs.
• Use the Stata memory command to determine how much memory is currently being used and how much is available. For example, when the following command is issued at the Stata dot prompt:
memory
The following report will be displayed:
Total memory 2,097,144 bytes 100.00%
overhead (pointers) 0 0.00%
data 0 0.00%
------------
data + overhead 0 0.00%
programs, saved results, etc. 352 0.02%
------------
Total 352 0.02%
Free 2,096,792 99.98%
This reports indicates that more than 2M bytes of memory are available for the current session, the default allocation in UNIX.
• Use the set memory command to allocate additional memory, if needed. The general syntax is:
set memory [memsize]
where memsize is the amount of memory requested. For example, issue the following command to request 4 MB of memory for your current Stata session:
set memory 4m
How to Reduce Memory Requirements
You may be able to reduce your memory requirements by saving your data more efficiently. You can use the compress command to reduce the amount of memory your data consumes.
This command does not create a compressed version of your file in the way that compression utilities do, such as zip or pkzip. Rather, the Stata command compress changes the data types used to store your variables such that each variable is stored optimally.
In numerous examples, memory reqirements have been reduced by as much as 85% using the compress command.
1. Check the size of the original file using the UNIX ls command. This command can be issued from within a Stata session:
ls -l myfile.dta
where myfile.dta is the name of your Stata file. The -l (long) option displays a long listing of the file that includes its number of bytes. In Windows, you can ude the DOS dir command:
dir myfile.dta
2. Open (use) your Stata file. For example:
use myfile
(Recall that the .dta file extension is assumed on the use command.)
3. issue the compress command:
compress
Stata will consider changing the data type for each of your variables. A message will be displayed if the storage type is changed.
4. You need to save the new version of the file in order for the changes to remain in effect. Issue the Stata save command followed either by a new name for your Stata file or by the replace operand. For example
save myfile2
will save a new Stata file named myfile2.dta. You will have the original file, myfile.dta plus the new file.
Alternatively, the following command will overwrite the original file:
save, replace
This will save the data is in Stata's work area as a file named myfile.dta, replacing your original data file. You should be certain that you want to overwrite your original Stata file before using the replace option.
5. Now for the fun part. Check the size of the new file using the UNIX ls command. This command can be issued from within a Stata session:
ls -l myfile2.dta
where myfile2.dta is the name of your Stata file. You may see a significant decrease in the size of your file. You need to compress your Stata file only once.
Examining your Data
Describe
Once you have the data in Stata, you will want to make sure that all the variables are there and that they are in the format you need. You can do this with the "describe" command. Describe, which can be abbreviated as simply "d," will provide basic information about the file and the variables. You don’t have to call the data into Stata to be able to describe it, though. The command:
d using myfile :
will accomplish this. This can be very useful if the the file is too large to fit into memory. Often, your dataset will be larger than the default and you will need to increase the amount of memory Stata uses. To do this, look at the "size" line in the describe output; this is the size of the file in bytes. Since 1,000,000 bytes is a megabyte, files larger than 4,000,000 will not be loaded into Stata unless you increase the memory with the "set memory" command:
set mem 10m
This example increases the memory to 10 megabytes. Make sure that you give yourself some extra memory so you can create new variables and/or add observations.
List
It is often useful to just look at the data without doing any kind of analysis. The "list" command, abbreiviated as "l" will let you do this. Simply giving the "l" command will display all of the data on your screen, if you specify certain variables in the command, then only those variables will be printed:
l :
l var1 var2 var3 - displays just var1, var2 and var3 :
In Stata for PC you can have a direct look at your data as if you were working with a spredsheet (e.g. Excel), by using the "Stata Editor". You enter the editor by clicking on the "Editor" button. :You can edit your data by highlighting the cell you want ot change, entering the new value for that cell, and clicking on "return". You can edit the variable name (and the variable label) by double clicking on the relevant column : .
Creating Variables
Stata can store data as either numbers or characters. Stata will allow you to do most analyses only on numeric data. Sometimes, when you use insheet or infile, a numeric variable gets read in as a character, or string variable. This may be due to a ‘space’ character or a ‘.’ in one of the cells. Since Stata allows you to do analyses only on numeric variables, you will need to convert them to numeric data.
There are different types of numeric variables - float, binary, double, long and int - the differences among them are simply how much space they take up in the file. In most cases, you will not need to concern yourself with these differences.
Encode
Use encode when the original variable is, indeed, a character variable (such as gender being coded as "m" and "f’") and you need numbers instead. The encode command does not produce dummy variables, it just assigns numbers to each group defined by the character variable. In this example, gender was the original character variable and sex is the new numeric variable:
encode gender, gen(sex) :
Generate
The "gen … real" command will create a numeric variable based on the original string variable. Use this if the original was defined as a character variable "by mistake." The following command will create a new, numeric variable var1n by converting the string variable, var1:
gen var1n=real(var1)
Often, you will need to create new variables based on the ones you have already. The two most common ways of creating new variables is by using "generate" and "egen." Here are some examples of gen:
gen total= var1n + var2 + var3
gen cumtot=sum(total)
gen id=_n
In the first example, we generate a new variable called "total" which is simply the addition of var1n , var2, and var3. In the next example, we create a cumulative sum of total (the value of cumtot for this observation is the sum of total for all previous observations). Finally, we create an "id" variable by simply setting it equal to "_n", which is Stata’s way of numbering the observations in a dataset. This id variable can be very useful when you need to sort your data in different ways (see sorting below). The variable "_n" can be used at any time, but you must remember that it refers to the number of the observation in the current order, not the original order. Here are some more examples of gen and _n:
gen total2=42.3 if _n==125
gen lagtot=total[_n-1]
In the first example we simply set total equal to a specific value for the first through the 125th observation, namely, observation number 125. In the second, we create what is called a "lagged" variable, lagtot, which is the value of total from the previous observation designated by the "[_n-1]." You can use any valid arithmetic expression inside the brackets. Here are two last examples:
gen price = total + lagtot if company=="IBM"
replace price = total + cumtot if company~="IBM"
Here, we generate a new variable, "price", which is the sum of total and lagtot, but only if the value of company is IBM. Note two things here: first is the double "=" after "if", and second, the use of the quotes around IBM. When you refer to string variables in equality expressions such as this, you must put the quotes around the value so Stata will not confuse it with a variable name. Had we left the quotes out, Stata would have thought we wanted to generate price for observations where the variable company equals the variable "IBM." In the second example we use "replace" instead of "gen." This reason for this is that in the first example, Stata creates values only for those observation fulfilling the "if" condition. All other observations get "missing" values. We want, then, to replace those missing values with something. The "~" in Stata means "not," and suffices as the extra "=."
Dummy variables
Sometimes we need to generate a "dummy" variable, or variables. Stata makes this very easy:
tab size, gen(dummy) :
Here, Stata will create a dummy variable for each value found in size. So, dummy1 = 1 if it is a big company, 0 otherwise; dummy2 = 1 if it is a medium company, 0 otherwise; dummy3 = 1 if it is a small company, 0 otherwise. (See a list and desc)
If you want a dummy variable to indicate only a particular size category:
gen d_med = (size==2)
Here Stata will create a dummy variable such that: d_med = 1 if it is a medium company, 0 otherwise. In general, using this command will make Stata create a dummy variable equal to 1 for each observation where the expression in brackets is true, and equal to zero otherwise.
Character Variables
Stata assumes that the variable you are creating is numeric unless you tell it otherwise. Sometimes, though, you will want to create a variable whose values are strings. Here’s how:
gen str5 name="John"
gen str10 team="Atlanta" in 1/5
As above, the values for string variables must be enclosed in quotes. You must also specify the largest number of characters, including spaces, that the variable will have in the "str#". In other words, if "John" is the longest value for the variable name, then you must specify name to have at least 4 characters. It’s fine if you specify more characters than the values will actually take, but if you specify less, you will either get an error message, or some of the characters will be cut off.
The "in" option simply tells Stata to apply the "gen" command to observations 1 to 5, inclusive, in the order they appear in the dataset. You can specify any consecutive range. The "in" option can also be used with numeric variables, as well as any other commands.
gen str3 yn="yes" if team=="Atlanta" & name=="john"
replace yn="no" if team=="New York" | team=="Boston"
Here we extend the use of the "if" option to include two expressions. As you may have guessed, the "&" means "and." The "|" means "or." One thing you must be wary of is that the values of character variables are case-sensitive, so "John" is not the same as "john."
gen day=substr(code,1,2)
The substr function can be helpful; it specifies certain parts of a character variable to be used. The first argument specifies the variable to be picked apart (code), the second argument is the starting character, and the last argument is the number of characters to be pulled out. In the first example, starting with the first character in the variable "code," two characters will be extracted.
Date Variables
Stata stores dates as the number of elapsed days since January 1, 1960. There are two ways to create elapsed dates, depending on how your original variables are stored. If you have separate variables for month, day and year then use the mdy() function to create an elapsed date variable. If your original dataset already contains a date variable, then use the date() function. Once you have created an elapsed date variable, you will probably want to format it, as described below.
• Use the mdy() function to create an elapsed date variable when your original data contains separate variables for month, day and year. The month, day and year variables must be numeric. For example, suppose you are working with these data.
month day year
7 11 1948
1 21 1952
11 2 1994
8 12 1993
Use the following Stata command to generate a new variable named mydate:
gen mydate = mdy(month,day,year)
where mydate is an elapsed date varible, mdy() is a Stata function, and month, day, and year are the names of the variables that contain data for month, day and year, respectively. The mdy() function can put three variables together to create a new date variable. The variables do not have to be called month, day, and year, but they do have to follow that order. One more caveat is that the year must be stored as a four-digit number, otherwise Stata will not know what century you want. You may need to do some more programming to add "1900" to the value.
• Use the date() function to create an elapsed date variable when your original dataset contains a date variable. For example, suppose you have a string variable, oldate, that represents the date of some event.
olddate
7-11-92
3-24-93
9-16-94
12-18-1996
Use the following Stata command to generate a new variable named newdate:
gen newdate = date(oldate,"mdy")
where newdate is an elapsed date varible; date() is a Stata function; olddate is the name of the variable that contains a date; and "mdy" is the format for reading the data (i.e., month, day and year).
Other formats are available for the date() function including "dmy" and "ymd". The format must match the order in which the values appear in the original date variable. You can use only one format to read a given date variable since the month, day and year components must appear in the same order for every occurance of the original date variable. However, you can represent the month, day and year components in a number of ways, for example:
somedate
7/11/92
Sept 24, 1993
9.16.94
December 18, 1996
The date() function can be used to read these data and create an elapsed date variable named newdate2:
gen newdate2 = date(somedate,"mdy")
• Use the format command to display elapsed dates as calendar dates. In the example given above, the elapsed date variable, mydate, has the following values, which represent the number of days before or after January 1, 1960.
month day year mydate
7 11 1948 -4191
1 21 1952 -2902
8 12 1993 12277
11 2 1994 12724
You can use the format command to display elapsed dates in a more customary way. For example:
format mydate %d
where mydate is an elapsed date variable and %d is the format which will be used to display values for that variable.
month day year mydate
7 11 1948 11jul48
1 21 1952 21jan52
8 12 1993 12aug93
11 2 1994 02nov94
Other formats are available to control the display of elapsed dates (see the manuals).
Here are some other useful functions for working with dates:
gen day = day(date)
gen weekday = dow(date)
gen month = month(date)
gen year = year(date)
All of these examples assume that the date variable is an elapsed date. The day() function returns the number of the day, ie, 23. The dow() returns a number from 0 to 7: 0 if it is a Sunday and 6 if it is a Saturday. month() returns a number from 1 to 12. year() returns the year as a four-digit number. :
Extended Generate (egen)
"egen", or "extended generate" is useful when you need a new variable that is the mean, median, etc. of another variable, for all observations or for groups of observations. Egen is also useful when you need to simply number groups of observations based on some classification variables. Here are some examples:
egen sumvar1 = sum(var1)
egen meanvar1= mean(var1), by(var3)
egen count = count(id), by(company)
egen group = group(month year)
In the first example, we simply create a variable whose value for each observation is equal to the sum of var1 for all observations (all observations will have the same value). The second example shows how to create a variable that is the mean of another variable for each group designated by var3 (all observations within a group will have the same value). The third example simply counts the number of non-missing id values in each company (this basically counts the number of observations you have for each company). The last example simply assigns a number to each of the groups created by the combination of month and year. :
There are many functions that can be used with "extended generate". To have a brief summary, type:
help egen
Variable and Value labels
Now that we’ve created all these new variables, we’ll want some way of keeping track of what each one is and what their values mean. We can do this by creating variable labels and value labels. These labels are not necessary in Stata, they just make the output easier to read. Variable labels correspond to the variable names, whereas value labels correspond to the different values a variable may have. Here’s how to create them:
label variable var1 "The first variable"
This assigns a label to the variable, var1, so whenever the variable name var1 is displayed on the screen, the description "The first variable" will be displayed as well.
In Stata for PC, variable names and variable labels will automatically appear in the "Variables" window, and you can enter a variable name in the "Stata Command" window simply by clicking on it in the "Variables" window. :
Variable labels must be 31 characters or less. Often, though, it’s more helpful to label the values a variable can take so you don’t have to memorize them or keep referring to a codebook.
label define grp 1 "Male" 2 "Female" 3 "N/A"
label values gender grp
Value labels correspond to the actual number or letters in the data. They are used so that the printout will show "Male," "Female," and "N/A" instead of "1," "2," and "3". First you have to "define" the label, then associate that label with a variable. Each label can be associated with more than one variable, so if there are several "yes/no/maybe" questions, you just have to define one value label and use it for all the questions. You may associate only one value label with a particular variable, however. :
You can check the list and content of all value labels by typing:
label list
Keep, Drop, and Rename
Well, now we’ve created many new variables and converted some old ones. Since we no longer need all of these variables, we’ll want to eliminate some of the ones we don’t really need and perhaps rename some of the ones we keep. We can either keep the variable we are interested in:
keep code date var4
or we can drop those we don't need, and rename some of those we keep:
drop var1 var2 var3
rename code date
Here we drop the variables, var1, var2 and var3. Be careful! Once they are dropped, they can’t be picked up again unless you clear the data and use it again. You can drop as many variables as you want in one command. The same is true for "keep". The second example renames the variable "code" to "date." You can rename only one variable at a time. You can also drop (or keep) observations:
drop if size==1
or
keep if size~=1
will drop all observations whose size is equal to 1. :
________________________________________
Log and Do Files
Log files
Log and do files are very useful. Logs keep a record of what commands you have issued and their results during your Stata session. Do files are good for long series of commands that may need to be "tweaked" to work properly. They are also necessary to replicate things that you have done on new or modified datasets.
You can create a log file with the command:
log using filename
Where filename is any name you wish to give the file. You will find it helpful to use names that will help you to remember what you did during that session. Stata will automatically append an extension of ".log" to the filename. By default, everything displayed on the screen will be recorded in the log file. You can give the file any name you like, but you should use names that will help you remember what analyses you did.
If you work for a long time, this file can get to be very large and have quite a bit of unnecessary output. So, if you are trying different things to "see what happens," you may want to start and stop the logging several times in one session. You can stop the logging with the command:
log close
If you turn the logging on and off during a session, you may need to use the append and/or replace options:
log using filename, append
log using filename, replace
The append option simply adds more information to the file, whereas the replace option erases anything that was already in the file. Be careful! It’s often helpful to put comments in the log file to help you remember why you did something. You can do this by simply preceding anything you type with a "*".
But, what happens if you forget to start a log file? There is a way to save what you have done by using the log file and the #review command. The #review command will list the last however many commands you have issued. This command can be used at any time and you can specify any number of commands to retrieve. After some minor editing and changing the extension to ".do", you can run it as a do file. Here’s an example:
log using iforgot - start a log file
#review 20 - list the last 20 commands
log close - close the log file
The file, iforgot.log, will have the last twenty commands you issued in it. You can then run this file as a do file with a new log started.
In Stata for PC you open a .log file by clicking on the "Log..." button. : If you want to open an existing .log file, you have to choose whether to overwrite it or to append the output to its existing content :. At this point a "Stata Log" window will pop up, and the contents of the .log file will be visualized there :. Once the "Stata Log" window is open, every command you entere "Stata Command" window will have its output saved in the log file. By clicking on the "Log..." buttom you can choose either to keep the log file open and suspend saving the output in it, or close the log file :. You can also print your log file by opening it and using the "File/Print Log" menu.
Do Files
A "do" file is a set of commands just as you would type them in one-by-one during a regular Stata session. Any command you use in Stata can be part of a do file. Do files are very useful, particularly when you have many commands to issue repeatedly, or to reproduce results with minor or no changes.
You can use any editor you wish to create your do files. If you use pine for email, then you already know how to use pico. The only thing you need to remember is that if a command is longer than one line, then you need to use the #delimiter command. In Stata, hitting the return key tells Stata to execute the command. In a .do file, the return key is at the end of every line, so you need a way of telling Stata that the command is longer than one line. Here is an example of a short do file we’ll call mydofile.do that you create in a text editor:
log using mydofile - start a log file
#delimiter ; - set the delimiter to ";"
use mydata; - open the data file
des; - describe the file
collapse (mean) var1 var2 var3 (meadian) medvar=var1, by(var4);
- collapse the data
save mydata2; - save the collapsed data
#delimiter cr - set the delimiter back to the "enter" key
clear - clear the data from memory
log close - stop logging
To run this file in Stata, you would simply issue the command:
do mydofile
You don’t even need to exit Stata to cretae or edit the do file. The command:
!pico myfile.do
will allow you to enter the text editor, pico, without exiting Stata. Once you save the file and exit from pico you are automatically retuned to Stata. You can use the ! character to tell Stata that you want this command issued to UNIX (or to DOS if you are working with Stata for PC in Windows), rather than Stata itself. Any command you would normally issue to UNIX (or to DOS)can be issued from within Stata this way. The "ls" (list files) and "cd" (change directory) commands do not have to be prefixed by the !.
Two other useful things are the Ctrl-r and Ctrl-c keys. You can use Ctrl-r to recall on the command line the commands you have issued. Sometimes the output from a command is very long and you don’t need to look at all of it; you can use Ctrl-c to stop the output.
In Stata for PC, run a .do file by using the "File/Do..." menu : . You can edit it with any normal text editor ( e.g. )
________________________________________
File Management
Once you have all of your variables in the format you want, you’ll need to get the entire file in a format that will make it easier to use. You can do this by sorting, appending, merging and collapsing.
Sort
"sort" puts the observations in a data set in a specific order. Some procedures require the file to be sorted before it can work. You can sort a file based on more than one variable. One thing you must be careful of, though, is that Stata will randomize the order of the observations within the variables used to sort. This is why creating the id variable mentioned earlier is important. With the id variable, you will always be able to go back to the original order and start over.
sort month :
sort month year
In Stata for PC you can sort data in the Stata Editor, by clicking on the variable you want to sort by, and then clicking the "Sort" button.:
Append
Sometimes, you have more than one file of data which you need to analyze. One case may be that you have two files with the same variables but different observations. The other case is when you have two files with the same observations, but different variables.
Use append when you simply want to add more observations, in other words you already have data for 1990, and now you want to include new observations from 1991 for the same variables. For example:
use class90
append using class91 :
will add the observations from the file, class91 (what Stata refers to as the "using" dataset), to the end of class90 (what Stata refers to as the "master" dataset). Any variables with different names in the two files will have missing values for the observations from the other dataset.
Merge
If you have two files that have the same observations, but different variables, then you’ll want to "merge" them so you can use all of the variables at once. When you merge datasets, you are adding new variables to existing observations rather than adding observations to existing variables. There are two basic kinds of merges, a one-to-one and a match.
A one-to-one merge simply takes the two files and puts them side-by-side, regardless of whether the observations in each dataset are in the same order. A simple one-to-one merge isn’t very common, and isn’t recommended, even if you think all the observations match. Sometimes there may be a "glitch" that can throw off the order of the data. It’s just as easy to do a match-merge, and you’ll be able to check that all the observations matched correctly.
There are some "rules" when doing a match-merge. First, each dataset must have a "key" variable by which the observations can be matched - social security number is a good example. Second, both datasets must be sorted by this key variable. You can use more than one key variable if you want, such as month and year. These key variables must be of the same type (string or numeric) in both datasets. By default, if a variable is present in both datasets, then the values in the master dataset will remain unchanged. If the variables in the using dataset have the same names as ones in the master dataset, but represent additional information, then you will need to rename them in one of the datasets before you can merge them.
use class90
sort date
merge date using class91 :
In this example we simply match observations using the variable "date" as the key.
Sometimes, you will want to update the data you have, namely, fill in missing values. Using the update option will accomplish this:
merge date using class91, update
Values in the master dataset will be changed only if they are missing.
In some instances, you may want to replace non-missing values in the master dataset. Use the replace option in addition to the update option to do this:
merge date using class91, update replace
"replace" will not work by itself, it must always be used with update. Stata will not, under any circumstances, change a non-missing value in the master dataset with a missing value from the using dataset. You must use the "replace" command discussed earlier to do this.
Stata will automatically create a variable called _merge which will indicate the results of the merge. Always check this variable to make sure that you got what you wanted. Here are what the possible values of _merge mean:
1 = Observations from the master dataset that did not match observations from the using dataset.
2 = Observations from the using dataset that did not match observations from the master dataset.
3 = Observations from both datasets that did match.
4 = Observations from both datasets that did match, missing values in the master dataset were updated.
5 = Observations from both datasets that matched, values in the master dataset disagree with those in the using dataset.
Usually, you will want all the observations to have a value of 3. Values of 4 occur only when you use the "update" option. Values of 5 occur only when you use the "update" and "replace" options. If you need to merge another dataset, you have to re-name or drop _merge first, other wise you will get a "_merge already defined" error. :
Collapse
"collapse" is used when you want to create a dataset containing the means sums, etc., of the various groups in the data. One example might be when you have one dataset of monthly data and another of yearly data and you need to analyze both sets of information together.
collapse (mean) var1 var2, by(month)
This will create a dataset of the means of var1 and var2 by month. If you have twelve months in your original dataset, then you will have twelve observations in the collapsed dataset. You must be careful, though, because Stata will compute the statistics on a variable-by-variable basis. If one variable has more missing observations than another, the means (and any other statistics you request) will be based on a different number of observations. This is not always an acceptable practice. To avoid this, you must use the "cw" options for "casewise deletion." This means that Stata will drop any observation that does not have data for both var1 AND var2, thereby ensuring that all the statistics will be based on the same number of observations.
collapse (mean) var1 var2, by(month) cw :
________________________________________
Basic Commands
Now that you have your data in a format you want, check it before doing any analyses. This can save you quite a bit of frustration later on.
Summarize
"sum", short for summarize, will give you the means, sd’s, etc. of the variables listed. If you don’t list any variables, it will give you the information for all numeric variables. If a variable you thought was numeric shows up as having 0 observations and a mean of 0, then, most likely, Stata still thinks it’s a character variable.
summ :
summ var1 var2
The "detail" option gives you additional information about the distribution of the variable.
summ var1 , detail :
Inspect
"inspect" is another easy way to eyeball the distribution of a variable.
inspect var1 :
Tabulate
"tab", short for tabulate, will produce frequency tables. By specifying two variables, you will get a crosstab. There are other options to get the row, column and cell percentages as well as chi-square and other statistics; check the manual.
tab var1
tab var1 var2
tab var1 var2, row col chisq :
"By" group processing
Sometimes you’ll want to run a command or analysis on different groups of observations. The "by variable:" subcommand is the same thing as running the command with separate "if" statements for each group. You must sort the data before you can use the "by:"
sort sex
by sex: summ var2
by sex: gen lag2=var2[_n-1] :
Sometimes, this can produce more output on the screen than you would care to look at. To suppress output to the screen, but not to the log file, use "quietly."
quietly by var1: gen lag2=var2[_n-1]
________________________________________
Fancy commands
Macro variables
Sometimes you need to use many variables the same way many times. One example would be if you want to run regressions on different dependent variables using the same set of independent variables. This can mean a lot of typing. One way around this is to create a macro variable. A macro variable is simply a variable that has as its value a particular string. This string can be anything you specify: a list of variables, a particular command, or whatever. Whenever Stata comes across the macro variable, it will interpret it to mean whatever string you set the variable to.
local macvar var1 var2 lagtot lag2
reg depvar `macvar’
In this example, when Stata "sees" the variable macvar in the regression command, it replaces `macvar’ with the string "att att2 itt itt2 date." Pay careful attention to the different type of quotation marks: the first quote in `macvar’ is the opening left quote usually found under the ~. The second quote is the closing right, or single quote usually found under the ".
For
Other times, you will want to perform the same command on several variables. Again, this can mean a lot of typing. The "for:" command can also be very useful in these situations. It can use several types of variable lists, and with a little ingenuity, can be very powerful. Here are a few examples:
for var1-var25: replace @=. if @==99
for var*: replace @=. if @=99
for 1-3, ltype(numeric): gen q@=0
for a b c, ltype(any): gen str2 @="x"
The @ in the for command represents the variable names in the list, one by one. In the first two examples, Stata would interpret the command as though you had typed replace var1=. if var1=99, replace var2=. If var2=99, and so on.
In any command, you can specify two kinds of lists:
1. var1-var25 or date-q25 indicates all variables between var1 and var25, inclusive, even if they do not begin with "var" as found when you run the describe command.
2. var* indicates all variables starting with "var."
in the third example, the command will create three new variables called q1, q2, and q3 and sets them all equal to zero. The last example creates three new variables called a, b, and c, and sets them all equal to "x."
________________________________________
Regression
Linear Regression Model
In Stata you can estimate Linear Regression Models using the "regress" command (shortcut reg). Here we suppose that you are familiar with linear regression models, and more in general with multiple regression analysis, and thus we don't provide any introduction to this subject.
Now, suppose your dependent variable is depvar and your independent variables are listesd in varlist. Then the command to run the linear regression is:
reg depvar varlist
The first variable after the regress command is always the dependent variable (or left-hand-side variable), and the following list gives the relevant independent variables (or right-hand-side variiables).
Suppose, for example, that you have a data set with individual observations on earnings, and you are interested in regressing the natural logarithm of average weekly earnings (lnw) on a sex dummy variable (sex: 0=male, 1=female), on a race dummy variable (race: 0=white, 1=non-white), on years of schooling (school), on age (age) and on a constant. Then you should type
reg lnw sex race school age :
Notice that Stata automatically adds the "constant" (_cons) to the list of independent variables. If you want to exclude it, you can use the option nocons:
reg lnw sex race school age, nocons
As for many other commands in Stata, the regress command can be also applied to a subset of your data set by specifying an if (or in) statement. Suppose that you want to run a regression only for the male population:
reg lnw race school age if sex==0 :
Several options are available with the regress command (see help regress). Among the most useful is the robust option, that allows you to run the regression using a "robust" estimator for the variance (specifically, the Huber/White/sandwich estimator). This option is mainly used when the residuals are thought to be heteroschedastic:
reg lnw sex race school age, robust
Another option that may be useful is the one that allows you to run Instrumental Variable regressions (or 2SLS). Suppose you have a dependent variable y1 and a list of independent variables x1 x2 x3; and let's suppose that you want to instrument x3 with z1, x1 and x2. Then you can simply run the (2SLS) regression by typing:
reg y1 x1 x2 x3 (z1 x1 x2)
and you will get the results of the second stage in the 2SLS procedure.
Logit, Probit
When you dependent variable is dichotomous (zero-one), you can run a Linear Probability Model (using the regress command) or you might be interested in running a Logit or a Probit regression. Here we assume that you are familiar with the Logit and Probit models, as well as with maximum likelihood estimation. As a quick reminder, the only difference between the Linear Probability Model (LPM), the Logit model and the Probit model is the assumption that you make about the probability distribution of your dependent variable. Let y be your dependent variable (dummy: zero-one), and let x and z be your independent variables. Then the general specification in all these three models is
Prob (y = 1) = F (bx + cz), Prob (y = 0) = 1 - F (bx + cz)
where b and c are the coefficeints of x and z, and F(.) is some function. The LPM assumes that F(a) = a. The Logit model assumes that F(.) is the logistic distribution function, while the Probit model assumes that F(.) to be the normal distribution function.
To run a Logit regression in Stata, you simply type:
logit depvar varlist
where depvar is your dependent (dichotomous) variable, and varlist is the list of independent variables that enter the model. To run a Probit model, the command is similar:
probit depvar varlist
Suppose that you are trying to estimate the effectiveness of a new trug being test for cancer patients. Your dependent variable is died (1 if patient died, 0 otherwise), and your independent variables are age (age of patient at start of experiment) and placebo (1 if drug is placebo). Then your logit and probit estimates will be: : .
Stata reports the maximum likelihood estimates for the original coefficients in both the probit and logit commands. You might be interested in other "coefficients". The dprobit command estimates a Probit model, but reports the change in the probability for an infinitesimal change in each independent variable instead of the coefficient for each independent variable. The logistic command estimates a Logit model but reports the odds ratios for each independent variable instead of the coefficients. After both these commands, you can obtain the original coefficients by simply typing (respectively) probit or logit :.
The logistic command also offers several "post-estimation" commands that can used to evaluate the goodness of fit of the model. These commands are entered after the estimation is performed, and are always referred to the "last" logistic estimation perfomed. See help logistic for a list.
Finally, all the above commands (as in regress) can be applied only to a subset of the observations, can be run without constant, and can generate estimates using robust standard errors.
Prediction
After all estimation commands (i.e. reg, logistic, logit, dprobit, probit) several "predicted" values can be computed. The most important are the "predicted values for the dependent variable" and the "predicted residuals". In order to obtain the predicted values for the dependent variable for any model you have to generate a new variable (call it yhat) that associate to each observation its predicted value based on the coefficients obtained from the estimation and based on the particular values of the independent variables for that observation. Suppose that you run a linear regression and you want to estimate the predicted (ln) average weekly wage for each individual: then you type
reg lnw sex race school age
predict yhat
then a new variable called yhat will be generated by Stata (if such a variable already exists you will get an error message, and you will have to choose another name for the predicted values) : . In the same way you can build a new variable (call it res) with the predicted "residual" for each observation:
reg lnw sex race school age
predict res, residual
Notice that after the logistic command, several other "predictions" are possible using the lpredict command (see help lpredict).
Testing Linear Hypothesis
Here we assume that you are familiar with the theory of hypothesis testing. After each estimation, for each independent variable Stata automatically provides a t-test (for linear regressions) and a z-test (for Logit or Probit models) on the null hypothesis that the "true" coefficient is equal to zero. However you can perform several other tests on linear hypothesis about the coefficients. Suppose you run a linear regression of (log) average weekly earnings on sex, race, schooling and age. You could test the hypothesis that the coefficient on race is equal to -.05 (i.e. earnings for non-whites are 5% lower than for whites):
reg lnw sex race school age
test race=-.05 :
You could also be interested in testing hypotheses on several variables. For example you could be interested in testing the null hypothesis that the coefficients on sex and race are jointly equal to zero, or that they sum up to one:
reg lnw sex race school age
test sex race - tests the null that they are jointly equal to zero
test sex + race == 1 :
If you are interested in non-linear hypothesis testing, you can look at help testnl .
BARRIERS TO CCS IMPLEMENTATION in AUSTRALIA
BARRIERS TO CCS IMPLEMENTATION in
I. INTRODUCTION
It is believed that in the absence of commercially successful CCS, some coal-mining and coal-based power-generating firms in
Many energy agencies have noted that global power generation will continue to rely on fossil fuels, but that carbon capture and storage is vital to that continued use (Mckee, 2006). However, the technique must be environmentally sustainable, cost-effective and capable of being broadly applied. There are a number of obstacles including: technology development and deployment, financing, legal-regulatory framework, capacity building, environmental assessment, public awareness and acceptance.
II. TECHNICAL BARRIERS
2.1 Low Concentration of CO2 Emitted From Power Stations
The purpose of CO2 capture technologies is to produce a concentrated stream of CO2 at high pressure that can be transported to a storage site (
In
2.2 Difficulties with Retro-Fitting
Once a base-load power station is built, very little can be done to reduce its emissions over its life, which could be from 25 to 40 years (
In the case of Callide Oxyfuel Project at
2.3 Loss of Efficiency
CO2 capture systems require significant amounts of energy. This additional energy reduces the efficiency of power plants, leading to increased fuel requirements, solid wastes and environmental impacts relative to the same type of base plant without capture (IPCC, 2005). Effectively, power plants with CCS require more fuel to generate each kilowatt-hour of electricity produced and most of this additional energy is the energy required for capture and compression of the CO2 (IPCC, 2005).
Other studies also suggest that the generating efficiency would be reduced by 10-15% based on current technology (CO2CRC, 2009). The International Energy Agency notes that the older power stations are not efficient (IEA, 2006). However, as more efficient plants with the availability of capture and replace many of the older-less-efficient plants, the net impact will be a reduction in emissions (
2.4 Advance Problems
CO2 is already being captured in the oil, gas, chemical and food industries. However, the existing capture technologies were not developed specifically for large scale carbon capture from power stations. To reduce emissions from a typical power plant by 75%, the equipment would need to be ten times larger. Yet, according to CO2CRC (2009), there have been no applications of CO2 capture technology in
The major challenge is to mount a project at the 500 MW scales which demonstrates all stages in the process from coal conversion, carbon capture, and transport, through to sequestration and long-term monitoring (HRSCSI, 2007). This raises logistic coordination and environmental and technical challenges that are not tested or resolved by small-scale demonstrations. The high capital cost of installing the huge post-combustion separation systems is also a major impediment to post-combustion capture of CO2 (
2.5 Poor Source to Sink Match for Some Major Emissions Regions
Large sources of CO2 are concentrated near major industrial and urban areas. Globally, many of these emission sources are within 300km of areas that potentially hold formations suitable for geological storage of CO2 (
In 2004, 65 Environmentally Sustainable Sites for CO2 Injection, (ESSCIs) were identified and their total risked storage capacity is 740 Gt CO2/year, while current emissions from stationary point sources in Australia are 309 Mt CO2/year (Baldwin, 2008). However, this estimate does not reflect commercial aspect where each ESSCI must be analyzed in terms of project specific economics and include the costs of compression, transport and injection (
2.6 No Integrated Full-Chain CCS Demonstrated
While each of the key CCS components of capture, transport and storage have been demonstrated in various industrial applications, they did not show the full chain capability, nor at the scale required to prove its application for large-scale power generation (MacGill et al, 2006). The major components of CCS are commercially available, yet, there is relatively little commercial experience with configuring all of these components into fully integrated CCS systems at the kinds of scales which would likely characterize a future deployment in
III. ECONOMIC COST of CCS
HRSCSI (2007) indicated that if CCS is not included, more expensive technologies will have to be utilized to reduce CO2 emissions. While, the IPCC (2005) estimated that, in the long term, including CCS in the range of mitigation strategies will reduce the cost of stabilizing CO2 by upwards of 30 per cent.
In the Australian context, ABARE (cited in HRSCSI, 2007) estimated that if early action of including CCS is taken into account,
Surprisingly, the Australian Government responded hesitantly (HRSCSI, 2007). This is due to the heterogeneous nature of the technical options available including capture, transport, storage; the variability of its application; the technical and financial complexity of integration; and the still largely speculative nature of the risk profiles. In addition, IPCC (2005) stated that there is still little experience in a fully integrated CCS system as it has still not been used in large-scale power plants.
IV. LEGAL ISSUES
The regulatory framework will need to cover both onshore sequestration and offshore sequestration (HRSCSI, 2007). Currently state and federal legislation only covers access and property rights of sites. A nationally consistent framework is required which comprises issues such as transport, injection, monitoring and financial liability through the stages of CCS (HRSCSI, 2007).
There has been reluctance on the part of Australia Government to share risk, particularly long-term liability (Cook, 2008). Therefore, the issue of long term liability determines whether there is a major inhibition of CCS or not. On the one hand, CSIRO and Chevron (cited in HRSCSI, 2007) proposed that liability must be shared by operators and responsibility should be handed to the government once the site has been closed. On the other hand, Greenpeace
V. ENVIRONMENTAL RISKS of CCS
The greatest environmental risk associated with CCS relates to the long term storage of the captured CO2 (HRSCSI, 2007). Leakage of CO2, either gradual or catastrophic leakage, could negate the initial environmental benefits of CCS and may also have harmful effects on human health. While there is some experience with geological storage of CO2 and natural gas for periods of approximately 10-20 years, evidence of long-term storage over many thousands of years has not been proven (TRUenergy cited in HRSCSI, 2007). However, CSIRO (cited in HRSCSI, 2007) opposed that the ongoing study has increased confidence in the viability of CO2 storage. In the proponent view, IPCC (2005) suggested that the environmental risks associated with CO2 capture and storage are low.
In the case of abrupt leakage, it could occur if the well seal at the point of storage failed, resulting in the release of sequestered CO2 (HRSCSI, 2007). However, evidence shows that if storage sites are carefully selected, the chances of a catastrophic leak would be minimal. Current projects, the Otway Demonstration Project, extend understanding of the scientific processes and risk minimization associated with the selection, sequestration and monitoring of CO2 in an Australian context (HRSCSI, 2007). In contrary, gradual leakage could occur as a result of incorrect site selection and inadequate preparation. This leakage would compromise the initial objective of removing the CO2 from the atmosphere (HRSCSI, 2007).
VI. PUBLIC PERCEPTION OVER CCS
The Australian Government (cited in HRSCSI, 2007) noted that the public is not well informed on CCS technology and its potential for climate change mitigation. The major public concern relates to potential leakage, consequent impact on the environment, and guidelines to secure public involvement through consultation processes. However, Friends of the Earth Australia insisted that public consultation for the Otway Basin Pilot Project is insufficient, where this is contradict to the CO2CRC report which claimed that extensive consultations preceded the announcement (HRSCSI, 2007). Thus, the community needs to be fully convinced about the long-term safety of storing large volumes of CO2 deep underground.
VII. CONCLUDING REMARK
CCS should be considered as a promising but still somewhat unproven option that potentially offers very significant abatement potential and good integration into the existing energy industry. However, its abatement is likely to come at a significant cost, and it is unlikely to be able to make a significant contribution for well over a decade. The actual implementation of CCS is likely to be lower than the economic potential due to factors such as environmental risks, the lack of a clear legal framework or public acceptance. In addition, there are many ways in which CO2 emissions can be reduced, however, most scenarios suggest that CCS will be only part of a suite of solutions.
REFERENCE
Baldwin, Stephanie. 2008. “Carbon Capture and Storage.” NSW Parliamentary Library Briefing Paper No 2/08. http://www.parliament.nsw.gov.au/prod/parlment/publications.nsf/0/959E4E67AFC1FB63CA25743F007EC318/$File/FINAL_CCS%202008.pdf. Accessed on 7 May 2009
Bradshaw, J., Allinson, G., Bradshaw, B.E., Nguyen, V., Rigg, A.J., Spencer, L., & Wilson, P. 2004.
CO2CRC. “Otway Project Overview.” http://www.co2crc.com.au/otway/. Accessed on 6 May 2009
Cook, J Peter. 2008. “Demonstration and Deployment of Carbon Dioxide Capture and Storage in Australia.” http://www.sciencedirect.com/. Accessed on 7 May 2009
Garnaut, R. 2008. “The Garnaut Climate Change Review.” Final Report. http://www.garnautreview.org.au/CA25734E0016A131/pages/draft-report. Accessed on 7 May 2009
House of Representatives Standing Committee on Science & Innovation. 2007. “Between a rock and a hard place: The science of geo-sequestration.” http://www.aph.gov.au/house/committee/scin/geosequestration/report/front.pdf. Accessed on 6 May 2009
IEA. 2006. “Carbon Capture and Storage.” http://www.iea.org/Textbase/techno/essentials.htm. Accessed on 6 May 2009
IPCC. 2005. “Special Report Carbon Dioxide Capture and Storage.” http://www.environment.gov.au/settlements/industry/ccs/publications/key-findings.html. Accessed on 8 May 2009
MacGill,
Massachusetts Institute of Technology (MIT). 2007. “The Future of Coal: Options for A Carbon-Constrained World.” http://web.mit.edu/coal/The_Future_of_Coal.pdf. Accessed on 6 May 2009
Mckee, N. Barbara. 2006. “Power Sector Approaches to Carbon Capture and Storage.” https://www.iea.org/Textbase/work/2006/enel/Session%203/McKee.pdf. Accessed on 7 May 2009
Saddler H., Reidy, C. & Passey, R. 2004. “Geo-sequestration: What it is and how much can it contribute to a sustainable energy policy for
Subscribe to:
Posts (Atom)
.jpg)