Using Stata 15.
Regressing an earnings model with American Community Survey data.
The natural log of earnings is the dependent variable.
Using vce(sdr).
Takes 5.5 hrs. to complete and the regression results, R-squared, joint significance of categorical indicators, and coefficient standard errors are all good.
margins is also working.
I went to predict level earnings and found that e(rmse) is not set. I need this post-regression statistic for the transformation to level earnings. e(rss) is not set either.
Is this the expected behavior? I’m confused because R-squared is reported in regression output.
Is it because I am using vce(sdr). If I used linearized standard errors would e(rmse) be set?
I can try regressing level earnings, but the regression takes so long I was hoping to get some insight before I choose the next step.
Thanks in advance.
Tom
Regressing an earnings model with American Community Survey data.
The natural log of earnings is the dependent variable.
Using vce(sdr).
Takes 5.5 hrs. to complete and the regression results, R-squared, joint significance of categorical indicators, and coefficient standard errors are all good.
margins is also working.
I went to predict level earnings and found that e(rmse) is not set. I need this post-regression statistic for the transformation to level earnings. e(rss) is not set either.
Is this the expected behavior? I’m confused because R-squared is reported in regression output.
Is it because I am using vce(sdr). If I used linearized standard errors would e(rmse) be set?
I can try regressing level earnings, but the regression takes so long I was hoping to get some insight before I choose the next step.
Thanks in advance.
Tom
