How can one deal with missing data in the glm? I am using 3 EVs that represent Low, Med, High arousal however some subjects don’t have ratings that reach the “high” criteria. I’ve entered these EVs as 3-column custom files and I noticed there is an option to choose “Empty (all zeros)”. I’ve considered the option of entering normalized scores, or the mean of the groups but these are not options that make much sense in terms of our hypotheses. If I don’t include subjects with missing data somehow, our N drops significantly. 

We are trying several options and would like to know if this is a useful one.

Thank you,
--