Dear STATAList forum,
I am doing a 3 model HLR (see final output below) and have transformed my outcome into the natural log to correct for skewness & kurtosis so it is approximately normal. I would like to show my results in their exponentiated form as it will be more easily interpretable. I searched and found this post about the "eform()" command and it does not work with the "hireg" command. I then thought that I could just use the "eform()" command on each regression model separately to get the exponentiated coefficients (is that correct thinking??). However, when I did that, the p-values became wayyyyy significant, which is a good thing, but it seems fishy to me (see output & code below). So my question is, what is the best way to exponentiate my coefficients in STATA? If I exponentiate my coefficients can I keep the same p-values from the original output and just use the exponentiated coefficients, standard error and CI?
I also found this post that seemed to address this issue for other commands (i.e. mlogit, logit), but did address regress.
I would appreciate any help!
With gratitude,
Jessie
Model 3:
Variables in Model: MilHis bachelors young white employed Enlisted HisVioCut MOSSocial ArmLife spirit01
Adding : OffPost evnt02 evnt07 hvchld3 spmhtmt01 spwkfamTotal
Source | SS df MS Number of obs = 289
-------------+---------------------------------- F(16, 272) = 6.53
Model | 7.72625871 16 .482891169 Prob > F = 0.0000
Residual | 20.1284812 272 .074001769 R-squared = 0.2774
-------------+---------------------------------- Adj R-squared = 0.2349
Total | 27.8547399 288 .096717847 Root MSE = .27203
------------------------------------------------------------------------------
Ln_MH3 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
MilHis | .0006318 .0356191 0.02 0.986 -.0694924 .070756
bachelors | .0406356 .0459031 0.89 0.377 -.0497349 .1310061
young | .0292034 .0375756 0.78 0.438 -.0447726 .1031795
white | -.01629 .0392448 -0.42 0.678 -.0935522 .0609721
employed | -.0178595 .0365116 -0.49 0.625 -.0897407 .0540218
Enlisted | .0394833 .0514528 0.77 0.444 -.0618131 .1407797
HisVioCut | .0518625 .0366613 1.41 0.158 -.0203135 .1240385
MOSSocial | -.0146549 .0055752 -2.63 0.009 -.025631 -.0036787
ArmLife | -.0144174 .0051494 -2.80 0.005 -.0245552 -.0042796
spirit01 | -.0313265 .0373038 -0.84 0.402 -.1047673 .0421143
OffPost | .0291096 .0365824 0.80 0.427 -.0429111 .1011302
evnt02 | -.0802444 .0415086 -1.93 0.054 -.1619634 .0014746
evnt07 | .1008607 .0409194 2.46 0.014 .0203016 .1814198
hvchld3 | .0028135 .0141164 0.20 0.842 -.0249777 .0306048
spmhtmt01 | .1539404 .0456291 3.37 0.001 .0641094 .2437715
spwkfamTotal | .0057239 .0022383 2.56 0.011 .0013172 .0101306
_cons | 3.219909 .1256884 25.62 0.000 2.972463 3.467354
------------------------------------------------------------------------------
R-Square Diff. Model 3 - Model 2 = 0.092 F(6,272) = 5.600 p = 0.000
Model R2 F(df) p R2 change F(df) change p
1: 0.014 0.688(6,293) 0.660
2: 0.185 6.482(10,285) 0.000 0.171 14.980(4,285) 0.000
3: 0.277 6.525(16,272) 0.000 0.092 5.600(6,272) 0.000
. regress Ln_MH3 MilHis bachelors young white employed Enlisted HisVioCut MOSSocial ArmLife spirit01 OffPost evnt02 evnt07 hvchld3 spmhtmt01 spwkfamTotal, noconstant eform (GMRation) robust
Linear regression Number of obs = 289
F(16, 273) = 781.45
Prob > F = 0.0000
R-squared = 0.9756
Root MSE = .50163
------------------------------------------------------------------------------
| Robust
Ln_MH3 | GMRation Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
MilHis | 1.021243 .0675565 0.32 0.751 .8965414 1.16329
bachelors | 1.245405 .1094339 2.50 0.013 1.047568 1.480604
young | 1.046638 .0731799 0.65 0.515 .9120446 1.201093
white | 1.231182 .0968868 2.64 0.009 1.054482 1.437491
employed | .9853398 .0664106 -0.22 0.827 .8629005 1.125152
Enlisted | 2.047147 .1779596 8.24 0.000 1.725139 2.429259
HisVioCut | 1.325456 .086041 4.34 0.000 1.166445 1.506144
MOSSocial | 1.038297 .0098824 3.95 0.000 1.019023 1.057936
ArmLife | 1.0387 .0084188 4.68 0.000 1.022257 1.055407
spirit01 | 1.010598 .0688219 0.15 0.877 .883799 1.15559
OffPost | 1.310207 .0877212 4.04 0.000 1.148408 1.494801
evnt02 | .9936841 .0676167 -0.09 0.926 .8690988 1.136129
evnt07 | 1.271028 .1021157 2.99 0.003 1.085086 1.488833
hvchld3 | 1.064069 .0257738 2.56 0.011 1.01452 1.116039
spmhtmt01 | 1.092585 .0941941 1.03 0.305 .922029 1.294691
spwkfamTotal | 1.036121 .0037785 9.73 0.000 1.028709 1.043587
------------------------------------------------------------------------------
I am doing a 3 model HLR (see final output below) and have transformed my outcome into the natural log to correct for skewness & kurtosis so it is approximately normal. I would like to show my results in their exponentiated form as it will be more easily interpretable. I searched and found this post about the "eform()" command and it does not work with the "hireg" command. I then thought that I could just use the "eform()" command on each regression model separately to get the exponentiated coefficients (is that correct thinking??). However, when I did that, the p-values became wayyyyy significant, which is a good thing, but it seems fishy to me (see output & code below). So my question is, what is the best way to exponentiate my coefficients in STATA? If I exponentiate my coefficients can I keep the same p-values from the original output and just use the exponentiated coefficients, standard error and CI?
I also found this post that seemed to address this issue for other commands (i.e. mlogit, logit), but did address regress.
I would appreciate any help!
With gratitude,
Jessie
Model 3:
Variables in Model: MilHis bachelors young white employed Enlisted HisVioCut MOSSocial ArmLife spirit01
Adding : OffPost evnt02 evnt07 hvchld3 spmhtmt01 spwkfamTotal
Source | SS df MS Number of obs = 289
-------------+---------------------------------- F(16, 272) = 6.53
Model | 7.72625871 16 .482891169 Prob > F = 0.0000
Residual | 20.1284812 272 .074001769 R-squared = 0.2774
-------------+---------------------------------- Adj R-squared = 0.2349
Total | 27.8547399 288 .096717847 Root MSE = .27203
------------------------------------------------------------------------------
Ln_MH3 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
MilHis | .0006318 .0356191 0.02 0.986 -.0694924 .070756
bachelors | .0406356 .0459031 0.89 0.377 -.0497349 .1310061
young | .0292034 .0375756 0.78 0.438 -.0447726 .1031795
white | -.01629 .0392448 -0.42 0.678 -.0935522 .0609721
employed | -.0178595 .0365116 -0.49 0.625 -.0897407 .0540218
Enlisted | .0394833 .0514528 0.77 0.444 -.0618131 .1407797
HisVioCut | .0518625 .0366613 1.41 0.158 -.0203135 .1240385
MOSSocial | -.0146549 .0055752 -2.63 0.009 -.025631 -.0036787
ArmLife | -.0144174 .0051494 -2.80 0.005 -.0245552 -.0042796
spirit01 | -.0313265 .0373038 -0.84 0.402 -.1047673 .0421143
OffPost | .0291096 .0365824 0.80 0.427 -.0429111 .1011302
evnt02 | -.0802444 .0415086 -1.93 0.054 -.1619634 .0014746
evnt07 | .1008607 .0409194 2.46 0.014 .0203016 .1814198
hvchld3 | .0028135 .0141164 0.20 0.842 -.0249777 .0306048
spmhtmt01 | .1539404 .0456291 3.37 0.001 .0641094 .2437715
spwkfamTotal | .0057239 .0022383 2.56 0.011 .0013172 .0101306
_cons | 3.219909 .1256884 25.62 0.000 2.972463 3.467354
------------------------------------------------------------------------------
R-Square Diff. Model 3 - Model 2 = 0.092 F(6,272) = 5.600 p = 0.000
Model R2 F(df) p R2 change F(df) change p
1: 0.014 0.688(6,293) 0.660
2: 0.185 6.482(10,285) 0.000 0.171 14.980(4,285) 0.000
3: 0.277 6.525(16,272) 0.000 0.092 5.600(6,272) 0.000
. regress Ln_MH3 MilHis bachelors young white employed Enlisted HisVioCut MOSSocial ArmLife spirit01 OffPost evnt02 evnt07 hvchld3 spmhtmt01 spwkfamTotal, noconstant eform (GMRation) robust
Linear regression Number of obs = 289
F(16, 273) = 781.45
Prob > F = 0.0000
R-squared = 0.9756
Root MSE = .50163
------------------------------------------------------------------------------
| Robust
Ln_MH3 | GMRation Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
MilHis | 1.021243 .0675565 0.32 0.751 .8965414 1.16329
bachelors | 1.245405 .1094339 2.50 0.013 1.047568 1.480604
young | 1.046638 .0731799 0.65 0.515 .9120446 1.201093
white | 1.231182 .0968868 2.64 0.009 1.054482 1.437491
employed | .9853398 .0664106 -0.22 0.827 .8629005 1.125152
Enlisted | 2.047147 .1779596 8.24 0.000 1.725139 2.429259
HisVioCut | 1.325456 .086041 4.34 0.000 1.166445 1.506144
MOSSocial | 1.038297 .0098824 3.95 0.000 1.019023 1.057936
ArmLife | 1.0387 .0084188 4.68 0.000 1.022257 1.055407
spirit01 | 1.010598 .0688219 0.15 0.877 .883799 1.15559
OffPost | 1.310207 .0877212 4.04 0.000 1.148408 1.494801
evnt02 | .9936841 .0676167 -0.09 0.926 .8690988 1.136129
evnt07 | 1.271028 .1021157 2.99 0.003 1.085086 1.488833
hvchld3 | 1.064069 .0257738 2.56 0.011 1.01452 1.116039
spmhtmt01 | 1.092585 .0941941 1.03 0.305 .922029 1.294691
spwkfamTotal | 1.036121 .0037785 9.73 0.000 1.028709 1.043587
------------------------------------------------------------------------------
Comment