Analytics Ninja and The Curse of Multicollinearity
Starring Data Monger
Starring Analytics Ninja
Far far away in a distant office cubical, Data Monger's rants can be heard yet again!
Well, I have these predictor variables with large P valuesgenerating highly questionable results. I just don’t know how to pick the bestpredictor for this model and it’s driving me crazy.
Holy Pie! These two predictors have large P values and they seem to be highly correlated. This cannot happen to me…I need to get the first iteration of this model out ASAP, or I am gonna lose this project.
Somewhere in the core, Analytics Ninja can hear the call for help!
I wish Analytics Ninja could come help!
Meanwhile Analytics Ninja: Help is on its way my friend!
OK DM, this is phenomenon is called Multicollinearity. Fortunately,it can be taken care of by looking at VIF statistics of call the predictorvariables. Have you heard of any of this?