White's test for linear regression

noetsi

Fortran must die
#1
I think the following performs White's test for heteroskedacity and generates WHITE robust SE, but I am not sure.

Code:
ODS graphics on;
PROC REG DATA=WORK.test2
		PLOTS(ONLY)=ALL
	;
	MODEL WEEKLYEARNINGS_CLO = WEEKLYEARNINGS_ACC
		/		SELECTION=NONE
		ACOV
		VIF SPEC
	;
RUN;
ODS graphics off;
QUIT;
 
Last edited:

hlsmith

Not a robit
#2
I believe to get sandwich estimators (White's), all you have to do is type in the options on the model line.




Code:
model y = x / white;
I could be wrong, though - I am too lazy to look it up.
 

noetsi

Fortran must die
#3
Last edited:

noetsi

Fortran must die
#4
I ran the WHITE test with SPEC and got...

The average covariance matrix for the SPEC test has been deemed singular which violates an assumption of the test. Use
caution when interpreting the results of the test.

Note entirely sure what that means :p The test was very significant although the residuals did not look that bad.
 

hlsmith

Not a robit
#5
Sounds like one of those matrix algebra thingy majiggies. When I read on the test a couple of days ago it mentioned that as a possible issue.
 

noetsi

Fortran must die
#6
I always hate when those matrix algebra majiggles come into play :mad:

I guessed it was that, but I am not certain what it means in practice. I guess I will ignore it :p