How to take difference in mean of two regression coefficients?

[Vijayshree Jayaraman]

Hi,

I have run two regressions using the following command:

regress violence rel_earn  m_decision_making  num_children pr_rshp edu_level-hus_edu_level if employed==1
eststo reg4a
regress violence rel_earn  m_decision_making  num_children pr_rshp edu_level-hus_edu_level if employed==0
eststo reg4b

After this, I used the suest command to generate the mean difference. However, the difference between the values of the coefficients is incorrect. Only the standard errors seem to change. The code that I have used:

suest reg4a reg4b 
test [reg4a_mean = reg4b_mean]
eststo t_test
esttab reg4a reg4b t_test using reg4.tex, replace

[Jeff Wooldridge]

Not really sure what the issue is here. You should show your output and then point to what you think is wrong. suest does not change the estimation; it simply allows you to obtain the covariance among estimators from different estimations.

[Jared Greathouse]

I don't understand. If one OLS coefficient is 10 and in another model is 15, I don't really understand what it even means, to take the difference in means between them.

The only time this question may make sense is when I put on my meta-analysis hat and think "Well, I GUESS in certain instances you can calculate the Cohen's d mean difference to get an effect size", but that doesn't seem to be the issue, or if it is, you haven't said.