Tests before running regression

[Thi Nguyen]

Hi everybody,

I intend to use panel data for my regression. However, which are tests necessary to check the valid data before I run regression model?

[Marcos Almeida]

Welcome to the Stata Forum / Statalist,

I didn't understand what you meant by "tests before running" regression. That said, if you have panel data, well, a panel data analysis will probably be the appropriate model, instead of OLS regression. Additionally, after the regression of a panel data model you do have in the output the test which compares it with the OLS regression. Hopefully that helps.

[Thi Nguyen]

Hi Marcos,

Thank for your answer.

I just wonder whether I have to check multicollinearity, heterokedasticity and autocorrelation

[Carlo Lazzaro]

Thi:
the tests you mantioned (which should be run not before but after the regression has been performed) are not directly available for -xtreg-.
However, if you have a N>T panel dataset you can invoke robust/cluster standard errors if you detect/suspect heteroskedasticity (you can have an idea of what's going on by visual inspecting the residual distribution using -qnorm-, -scatter- or -histogram-) and/or autocorrelation (that can be detected via -pwcorr- looping over the residuals); multicollinearity can be investigated via -estat vce, corr-.