This FAQ illustrates the estout command that makes regression tables in a format that is commonly used in journal articles. The estout command was written by Ben Jann of ETH Zurich. You can download estout from within Stata by typing search estout (see How can I use the search command to search for programs and get additional help? for more information about using search).
Let’s illustrate use of the estout command using the high school and beyond data file.
use https://stats.idre.ucla.edu/stat/stata/notes/hsb2, clear (highschool and beyond (200 cases))
We will run 3 regression models predicting the variable read. The first model will predict from the variables female and write; the second model will predict from female, write and math; and the third model will predict from female, write, math, science and socst. After each regress we will run an estimates store command. We will then use estout to create a single table that will summarize these models side by side.
regress read female write Source | SS df MS Number of obs = 200 -------------+------------------------------ F( 2, 197) = 66.11 Model | 8401.94189 2 4200.97094 Prob > F = 0.0000 Residual | 12517.4781 197 63.540498 R-squared = 0.4016 -------------+------------------------------ Adj R-squared = 0.3956 Total | 20919.42 199 105.122714 Root MSE = 7.9712 ------------------------------------------------------------------------------ read | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- female | -4.532084 1.171072 -3.87 0.000 -6.84153 -2.222637 write | .7067537 .0616783 11.46 0.000 .5851192 .8283882 _cons | 17.40106 3.202315 5.43 0.000 11.08584 23.71628 ------------------------------------------------------------------------------ estimates store m1, title(Model 1) regress read female write math Source | SS df MS Number of obs = 200 -------------+------------------------------ F( 3, 196) = 68.34 Model | 10695.1896 3 3565.06321 Prob > F = 0.0000 Residual | 10224.2304 196 52.1644406 R-squared = 0.5113 -------------+------------------------------ Adj R-squared = 0.5038 Total | 20919.42 199 105.122714 Root MSE = 7.2225 ------------------------------------------------------------------------------ read | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- female | -2.739657 1.09497 -2.50 0.013 -4.899092 -.5802214 write | .3924361 .0732832 5.36 0.000 .2479114 .5369609 math | .4753659 .0716952 6.63 0.000 .3339729 .6167589 _cons | 7.986659 3.230313 2.47 0.014 1.616025 14.35729 ------------------------------------------------------------------------------ estimates store m2, title(Model 2) regress read female write math science socst Source | SS df MS Number of obs = 200 -------------+------------------------------ F( 5, 194) = 56.29 Model | 12383.6535 5 2476.7307 Prob > F = 0.0000 Residual | 8535.76652 194 43.9987965 R-squared = 0.5920 -------------+------------------------------ Adj R-squared = 0.5815 Total | 20919.42 199 105.122714 Root MSE = 6.6332 ------------------------------------------------------------------------------ read | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- female | -1.328513 1.046793 -1.27 0.206 -3.393068 .7360424 write | .1503085 .0778554 1.93 0.055 -.0032431 .3038601 math | .2934723 .0728471 4.03 0.000 .1497983 .4371463 science | .2508791 .0667312 3.76 0.000 .1192673 .382491 socst | .2694578 .0574134 4.69 0.000 .1562232 .3826923 _cons | 2.44264 3.107255 0.79 0.433 -3.685698 8.570977 ------------------------------------------------------------------------------ estimates store m3, title(Model 3) estout m1 m2 m3 --------------------------------------------------- m1 m2 m3 b b b --------------------------------------------------- female -4.532084 -2.739657 -1.328513 write .7067537 .3924361 .1503085 math .4753659 .2934723 science .2508791 socst .2694578 _cons 17.40106 7.986659 2.44264 ---------------------------------------------------
Now we have a perfectly fine table that just includes the regression coefficients. We will modify the estout command to add standard errors and stars for statistical significance. We will also format the output so that coefficients will have three decimal places and the standard errors to two decimal places. Note, the par option for “se” places parentheses around the standard error.
estout m1 m2 m3, cells(b(star fmt(3)) se(par fmt(2))) ------------------------------------------------------------ m1 m2 m3 b/se b/se b/se ------------------------------------------------------------ female -4.532*** -2.740* -1.329 (1.17) (1.09) (1.05) write 0.707*** 0.392*** 0.150 (0.06) (0.07) (0.08) math 0.475*** 0.293*** (0.07) (0.07) science 0.251*** (0.07) socst 0.269*** (0.06) _cons 17.401*** 7.987* 2.443 (3.20) (3.23) (3.11) ------------------------------------------------------------
The table is better now, but it can be improved further by putting the model names above the columns, adding a legend and by changing the label for “_cons” to “constant.”
estout m1 m2 m3, cells(b(star fmt(3)) se(par fmt(2))) /// legend label varlabels(_cons Constant) -------------------------------------------------------------------- Model 1 Model 2 Model 3 b/se b/se b/se -------------------------------------------------------------------- female -4.532*** -2.740* -1.329 (1.17) (1.09) (1.05) writing score 0.707*** 0.392*** 0.150 (0.06) (0.07) (0.08) math score 0.475*** 0.293*** (0.07) (0.07) science score 0.251*** (0.07) social studies score 0.269*** (0.06) constant 17.401*** 7.987* 2.443 (3.20) (3.23) (3.11) -------------------------------------------------------------------- * p<0.05, ** p<0.01, *** p<0.001
Next, we want to add some things to the table, like R-squared, residual degrees of
freedom and BIC. Stata has special names for each of these ancillary statistics,
"r2" is the name for R-squared, "df_r" for residual degrees of freedom and "bic"
for the BIC. You can get the names of these items from the ereturn list and
from the
estout m1 m2 m3, cells(b(star fmt(3)) se(par fmt(2))) /// legend label varlabels(_cons constant) /// stats(r2 df_r bic) -------------------------------------------------------------------- Model 1 Model 2 Model 3 b/se b/se b/se -------------------------------------------------------------------- female -4.532*** -2.740* -1.329 (1.17) (1.09) (1.05) writing score 0.707*** 0.392*** 0.150 (0.06) (0.07) (0.08) math score 0.475*** 0.293*** (0.07) (0.07) science score 0.251*** (0.07) social studies score 0.269*** (0.06) constant 17.401*** 7.987* 2.443 (3.20) (3.23) (3.11) -------------------------------------------------------------------- r2 0.402 0.511 0.592 df_r 197.000 196.000 194.000 bic 1410.783 1375.608 1350.106 -------------------------------------------------------------------- * p<0.05, ** p<0.01, *** p<0.001
Okay, we're almost done. We just need to clean up the lower part of the table giving each of the items a better label and adjusting the number of decimal places for each of the items.
estout m1 m2 m3, cells(b(star fmt(3)) se(par fmt(2))) /// legend label varlabels(_cons constant) /// stats(r2 df_r bic, fmt(3 0 1) label(R-sqr dfres BIC)) -------------------------------------------------------------------- Model 1 Model 2 Model 3 b/se b/se b/se -------------------------------------------------------------------- female -4.532*** -2.740* -1.329 (1.17) (1.09) (1.05) writing score 0.707*** 0.392*** 0.150 (0.06) (0.07) (0.08) math score 0.475*** 0.293*** (0.07) (0.07) science score 0.251*** (0.07) social studies score 0.269*** (0.06) constant 17.401*** 7.987* 2.443 (3.20) (3.23) (3.11) -------------------------------------------------------------------- R-sqr 0.402 0.511 0.592 dfres 197 196 194 BIC 1410.8 1375.6 1350.1 -------------------------------------------------------------------- * p<0.05, ** p<0.01, *** p<0.001
We now have a table that's acceptable for publication in many journals. Of course, each periodical defines its own formats. Fortunately, estout is very flexible and has many options that will adapt to almost any periodical's requirements.