6.11 Collinearity

Collinearity (see Section 5.20) is a characteristic of the predictors, not the outcome, and so can be a problem not only in linear regression, but in all regression models. For logistic regression, diagnose collinearity in the same way as for MLR, using VIFs or aGSIFs.

Example 6.3 (continued): Evaluate collinearity in the adjusted model without an interaction. The aGSIFs are all small so we conclude there is no problem with collinearity.

car::vif(fit.ex6.3.adj)

##                 GVIF Df GVIF^(1/(2*Df))
## alc_agefirst   1.044  1           1.022
## demog_age_cat6 1.094  4           1.011
## demog_sex      1.038  1           1.019
## demog_income   1.086  3           1.014