So I need to run a Poisson regression model to see if an additional variable provides lift. This variable is a binary flag.
The issue is the variable is heavily skewed. Maybe 1% of all records have this flag but a univariate look shows those with this flag cost our company more relative to those who don’t. I’d like to use and already established model that already controls for all other variables we look at with the addition of this flag.
Is there any tips on using a variable that is severely lopsided on way?