- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Posted 04-27-2010 04:38 PM
(1459 views)
When evaluating simple or multiple linear regression models using SAS/LAB, an analysis of heteroscedasticity is provided. One gets a comment like the following:
"There is statistical (or "strong statistical" or "weak statistical") evidence that [dependent variable] variation is dependent on the value of [independent variable]. A transformation of [dep var] might be useful in reducing non-constant variance."
Or
" There is no (or "not much") statistical evidence that [dep var] variation is dependent on the value of [indep var]."
Does anyone know what SAS/LAB uses as "statistical evidence" for determining heteroscedasticity?? The help for SAS/LAB is practically non-existent and I can't find anything documenting the procedures used in SAS/LAB model evaluation.
Thanks for any hints,
Frances
"There is statistical (or "strong statistical" or "weak statistical") evidence that [dependent variable] variation is dependent on the value of [independent variable]. A transformation of [dep var] might be useful in reducing non-constant variance."
Or
" There is no (or "not much") statistical evidence that [dep var] variation is dependent on the value of [indep var]."
Does anyone know what SAS/LAB uses as "statistical evidence" for determining heteroscedasticity?? The help for SAS/LAB is practically non-existent and I can't find anything documenting the procedures used in SAS/LAB model evaluation.
Thanks for any hints,
Frances
1 REPLY 1
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Ok, I think I found the answer to my question and am posting it here for those that may be interested.
From the SAS-L discussion group (someone had asked the same question):
One person said SAS/LAB uses Levene's test.
Also found (eventually) in the help:
The method SAS/LAB software uses for testing the assumption of equal within-group variances is based on Levene's test, as modified by Brown \& Forsythe (1974).
I'm assuming the Leven's test is what is used in the multiple linear regression case too.
From the SAS-L discussion group (someone had asked the same question):
One person said SAS/LAB uses Levene's test.
Also found (eventually) in the help:
The method SAS/LAB software uses for testing the assumption of equal within-group variances is based on Levene's test, as modified by Brown \& Forsythe (1974).
I'm assuming the Leven's test is what is used in the multiple linear regression case too.