We quickly become very accustomed to reading and interpreting regression analyses from our computer output. Regression analyses as reported in journals and other publications often look very different. Here are the results of a regression a typical analysis as produced by Stata.
xi: regress write female read socst i.prog
i.prog _Iprog_1-3 (naturally coded; _Iprog_1 omitted)
Source | SS df MS Number of obs = 200
-------------+------------------------------ F( 5, 194) = 42.68
Model | 9364.97492 5 1872.99498 Prob > F = 0.0000
Residual | 8513.90008 194 43.8860829 R-squared = 0.5238
-------------+------------------------------ Adj R-squared = 0.5115
Total | 17878.875 199 89.843593 Root MSE = 6.6247
------------------------------------------------------------------------------
write | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | 4.916003 .9478479 5.19 0.000 3.046593 6.785412
read | .3427226 .060063 5.71 0.000 .2242624 .4611829
socst | .2685807 .0584637 4.59 0.000 .1532746 .3838868
_Iprog_2 | .9975073 1.233344 0.81 0.420 -1.434978 3.429992
_Iprog_3 | -1.888857 1.389299 -1.36 0.176 -4.628927 .8512126
_cons | 18.06893 3.068803 5.89 0.000 12.01643 24.12143
------------------------------------------------------------------------------
Publication tables usually have only the regression coefficient and standard errors along
with some fit statistics. In Stata, the estimates table command can be used to create
publication type tables. Unfortunately, the star option is not
allowed with the standard errors. Below are two examples:
estimates store m1
estimates table m1, stats(N r2 r2_a) b(%6.3f) star
---------------------------
Variable | m1
-------------+-------------
female | 4.916***
read | 0.343***
socst | 0.269***
_Iprog_2 | 0.998
_Iprog_3 | -1.889
_cons | 18.069***
-------------+-------------
N | 200.000
r2 | 0.524
r2_a | 0.512
---------------------------
legend: * p<0.05; ** p<0.01; *** p<0.001
estimates table m1, stats(N r2 r2_a) se b(%6.3f) se(%6.3f)
------------------------
Variable | m1
-------------+----------
female | 4.916
| 0.948
read | 0.343
| 0.060
socst | 0.269
| 0.058
_Iprog_2 | 0.998
| 1.233
_Iprog_3 | -1.889
| 1.389
_cons | 18.069
| 3.069
-------------+----------
N | 200.000
r2 | 0.524
r2_a | 0.512
------------------------
legend: b/se
Alternatively, you can use the outreg command (findit outreg) to produce an ASCII
table of the results. Note: You will usually need to add spaces manually to get the columns to
line up correctly.
outreg using table1
type table1.out
writing score
female 4.916
(5.19)**
reading score 0.343
(5.71)**
social studies score 0.269
(4.59)**
prog==2 0.998
(0.81)
prog==3 -1.889
(1.36)
Constant 18.069
(5.89)**
Observations 200
R-squared 0.52
Absolute value of t statistics in parentheses
* significant at 5%; ** significant at 1%
Now, let's run a second regression model with interaction and display the results of both
analyses in the same table.
xi: regress write female read i.prog*socst
i.prog _Iprog_1-3 (naturally coded; _Iprog_1 omitted)
i.prog*socst _IproXsocst_# (coded as above)
Source | SS df MS Number of obs = 200
-------------+------------------------------ F( 7, 192) = 31.84
Model | 9604.82727 7 1372.11818 Prob > F = 0.0000
Residual | 8274.04773 192 43.0939986 R-squared = 0.5372
-------------+------------------------------ Adj R-squared = 0.5203
Total | 17878.875 199 89.843593 Root MSE = 6.5646
------------------------------------------------------------------------------
write | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
female | 4.961672 .9408011 5.27 0.000 3.106039 6.817305
read | .3478027 .0599996 5.80 0.000 .2294597 .4661457
_Iprog_2 | 16.75438 6.793211 2.47 0.015 3.355472 30.15328
_Iprog_3 | 8.74215 6.838349 1.28 0.203 -4.745785 22.23009
socst | .475497 .1110706 4.28 0.000 .2564217 .6945723
_IproXsocs~2 | -.300757 .1274851 -2.36 0.019 -.5522081 -.0493059
_IproXsocs~3 | -.210099 .1386112 -1.52 0.131 -.4834953 .0632973
_cons | 7.321844 5.673858 1.29 0.198 -3.869253 18.51294
------------------------------------------------------------------------------
estimates store m2
estimates table m1 m2, stats(N r2 r2_a) b(%6.3f) star
----------------------------------------
Variable | m1 m2
-------------+--------------------------
female | 4.916*** 4.962***
read | 0.343*** 0.348***
socst | 0.269*** 0.475***
_Iprog_2 | 0.998 16.754*
_Iprog_3 | -1.889 8.742
_IproXsocs~2 | -0.301*
_IproXsocs~3 | -0.210
_cons | 18.069*** 7.322
-------------+--------------------------
N | 200.000 200.000
r2 | 0.524 0.537
r2_a | 0.512 0.520
----------------------------------------
legend: * p<0.05; ** p<0.01; *** p<0.001
estimates table m1 m2, stats(N r2 r2_a) se b(%6.3f) se(%6.3f)
----------------------------------
Variable | m1 m2
-------------+--------------------
female | 4.916 4.962
| 0.948 0.941
read | 0.343 0.348
| 0.060 0.060
socst | 0.269 0.475
| 0.058 0.111
_Iprog_2 | 0.998 16.754
| 1.233 6.793
_Iprog_3 | -1.889 8.742
| 1.389 6.838
_IproXsocs~2 | -0.301
| 0.127
_IproXsocs~3 | -0.210
| 0.139
_cons | 18.069 7.322
| 3.069 5.674
-------------+--------------------
N | 200.000 200.000
r2 | 0.524 0.537
r2_a | 0.512 0.520
----------------------------------
legend: b/se
outreg using table1, append
type table1.out
(1) (2)
writing score writing score
female 4.916 4.962
(5.19)** (5.27)**
reading score 0.343 0.348
(5.71)** (5.80)**
social studies score 0.269 0.475
(4.59)** (4.28)**
prog==2 0.998 16.754
(0.81) (2.47)*
prog==3 -1.889 8.742
(1.36) (1.28)
(prog==2)*socst -0.301
(2.36)*
(prog==3)*socst -0.210
(1.52)
Constant 18.069 7.322
(5.89)** (1.29)
Observations 200 200
R-squared 0.52 0.54
Absolute value of t statistics in parentheses
* significant at 5%; ** significant at 1%
Phil Ender, 5Jan98