Dear All, I am trying to do a logistic regression based on a big file (marketing campaign) with a lot of variables: 70,000 cases and circa 3,000 respondents (dependent variable coded 0 & 1). Basically, the results I get are somehow not what I would expect : The SPSS program actually predicts overall over 95% of my records correctly but the model predicts all the cases to be non-respondent (0). Could someone tell me whether it is because I have too many cases compared to respondents ? If so should I take a random sample of the non-respondents and how many ? Thank you for your help in advance, Regards TD %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%