Hi folks,
Insights please!
I am running logit model with complex survey data. All the independent variables categorical variables. Three of IVs, two of which with six categories each and one with 11 categories, seem to be highly inter-related. I suspect there are multiple collinearities among them. Hoping to address these to some extent, I tried to include some interaction terms. But I am still in dilemma to arbitrarily drop one of them. I found it is harder to diagnose collinearities among categorical variables with multiple categories. I even tried to calculate tolerance for each variable by running mlogit (treating each of them as DV turn by turn) but it did not work.
Q: How can I run diagnostics if there are collinearities among the three variables?
Thank you in advance!
Insights please!
I am running logit model with complex survey data. All the independent variables categorical variables. Three of IVs, two of which with six categories each and one with 11 categories, seem to be highly inter-related. I suspect there are multiple collinearities among them. Hoping to address these to some extent, I tried to include some interaction terms. But I am still in dilemma to arbitrarily drop one of them. I found it is harder to diagnose collinearities among categorical variables with multiple categories. I even tried to calculate tolerance for each variable by running mlogit (treating each of them as DV turn by turn) but it did not work.
Q: How can I run diagnostics if there are collinearities among the three variables?
Thank you in advance!
Comment