Heteroscedasticity Adjusted Standard Errors

Youtube Link: http://www.youtube.com/watch?v=paYqAb2npTc


See also: testing heteroscedasticity statistically


In a previous post, I discussed two procedures that can be used to test for heteroskedasticity from a statistical significance stand-point. In the presence of heterokedasticity, one should probably assume that the standard errors associated with the beta weights of the regression model are going to be inaccurate (likely underestimated). Fortunately, regression standard errors can be adjusted, when the assumption of homoskedasticity if violated. Unfortunately, few statistical packages offer this procedure as a matter of course. It must be implemented using a macro.

I found such a macro developed by Andrew F Hayes, which can be found here:


The macro was published in the following reference:

Hayes, A. F., & Cai, L. (2007). Using heteroscedasticity-consistent standard error estimators
in OLS regression: An introduction and software implementation. Behavior Research Methods, 39(4), 709-722.

Once the macro has been run (only once), you need to run the following syntax:

HCREG dv = "variable_name"
/iv = "variable_name1" "variable_name2", etc.
/const = 1
/method = 3
/covmat = 1  
/test = 1