3

I'm slightly confused as how to interpret the answers Stata is feeding me from the White's test. I am running two regressions:

Regression 1

White's test for Ho: homoskedasticity
         against Ha: unrestricted heteroskedasticity

     chi2(65)     =     51.59
     Prob > chi2  =    0.8866

Regression 2

White's test for Ho: homoskedasticity
         against Ha: unrestricted heteroskedasticity

         chi2(65)     =     70.49
         Prob > chi2  =    0.2991

In my opinion there is not enough evidence to conclude heteroskedasticity and the errors are homoscedastic. Am I wrong?

Nick Cox
  • 56,404
  • 8
  • 127
  • 185
Silver
  • 31
  • 1
  • 1
  • 2

1 Answers1

3

The interpretation is the same as for any other test. The $P$-values imply that you fail to reject the null hypothesis in each case.

I wouldn't use the wording that there is homoskedasticity; if anything, there is always heteroskedasticity, just that sometimes (often, even) there is not enough to worry about.

Plotting residuals versus fitted is as or more useful than this kind of test, to see what is going on with your regression, but if you are an economist your texts may omit to tell you that. (In Stata, that's rvfplot after regress. See the post-estimation help for the command you used.)

Nick Cox
  • 56,404
  • 8
  • 127
  • 185