Missing Values For Independent Variable in Logistic/Probit Regression
Posted: Fri Mar 31, 2017 8:46 am
I'm building a logistic/probit regression model to predict the propensity for a customer to convert (to some measure of sales). One of the independent variables is the time lag (in mins/days) between when someone first became a customer and when they next logged back into the website, e.g. 5 mins, 2 hours, 3 days, 15 day etc. People that never came back to the site will not have this ppoulated. So what value do I assign to those people that never came back to the site? Could I just give the max value from the valid cases/customers for this variable or would that incorrectly skew the results/coefficients? I don't want to exclude these people from the modelling (as I have done to date) because they have valid data on other variables in the model with high predictive power so I need to include these people! What are my options? Many Thanks!