04-03-2017 10:17 AM
Is there any easy way I can get OLS standard errors using "proc surveyreg"?
I know that proc surveyreg is a wonderful procedure to get robust standard errors that I cannot easily obtain from proc reg. What I am wondering now is exactly the other way around: getting proc reg standard errors from proc surveyreg.
Is there any options embeded in proc surveyreg that generates proc reg standard errors? Or, is there any simple manipulation of surveyreg outputs from which I can develop proc reg standards errors?
04-03-2017 10:29 AM
If you don't use any of the sample design or population information options or statements I would expect to get pretty close the to results from Proc Reg.
Is there a problem with running proc reg???
04-03-2017 11:14 AM
My sample is mostly panel data, with id's for an individual and dates for the individual's particular date. So, there are mutiple obs (dates) for an id. This means I need to cluster standard errors by id, and use fixed effects for dates.
So, this data suit for proc surveyreg with the "cluster" and "class" command. But often I need to see "OLS" standard errors. "proc reg" is not good in this case, since generating fixed effects is not much convenient with proc reg. "proc glm" neigher, since "proc glm" does not print adjusted R-squares.
I know that I can manipulate "poc glm" outputs or manipluate data for "proc reg" in order to get what I want. But just wondering if there is any easy way to get what I want from "proc surveyreg" and reduce my work.
Thanks for your answering, Ballardw.