标题: Mock Exam : PM : Q42: Multicollinearity with t-test and F-te [打印本页] 作者: dmar 时间: 2013-4-11 17:17 标题: Mock Exam : PM : Q42: Multicollinearity with t-test and F-te
Okay the F test declares significance and the IDVs’ T-tests p-values prove the significance of each IDV as well.
Yet, correlation between variables is the reason to rule out Multicollinearity.
So WHAT is the true test to rule out Multicollinearity ?
F test & test comparisons giving results of significance in consonance.
or Low Correlation between the IDVs ?作者: chetan86 时间: 2013-4-11 17:22
There isn’t a “test” for Multicollinearity to my knowledge in the same way as a DW or Breutsche Pagen (sic).
Just look for high R squared and insignificant t test.作者: transferpricing 时间: 2013-4-11 17:30
exactly. no formal test, but high F (model significance) and low t stats are a good hint that you might have it作者: Daniel1985 时间: 2013-4-11 17:34
Also, as what we saw on one of the samples/mocks, think about the definition of multicollinearity, which is correlation of the independent variables.
the vignette may say something about the correlation, but not actually show us R^2
