R square in -reghdfe-

[Shi Y. Yan]

I use -reghdfe- to run a fixed effect regression. I get the results with R squared, adjusted R squared and within R squared. Which one should I report in the paper? within R square is much smaller than the other two. (When I use -areg- to run fixed effect regression, there is no within R square)


later, I use -reghdfe- to run IV (2sls). It has Centered R2 and uncentered R2. what does it mean?


Can anyone help with the questions? Thanks

[Sergio Correia]

Sorry for not replying earlier. There are four possible R2 options to choose (overall, within, adjusted, and adjusted within). You can see them with "ereturn list", even though you only see three on the table.

IMHO you should use the within or adjusted within, but those tend to give you very very small R2s, so using many people tend to use the R2 overall. I do not know of a best practice so far, even though I've thought about that a bit: http://www.statalist.org/forums/foru...sted-within-r2

About the IV, check the help for ivreg2. Usually the centered one is the one you care, but in any case R2s are useless on IV regressions (see the Stata.com FAQ about that: http://www.stata.com/support/faqs/st...least-squares/ )

[Shi Y. Yan]

Thanks for all your helps, Sergio.