1

I am running a logistic regression in SAS where the outcome variable is a binary on sale success (IVs are: location of the customer, Vendor, handling time, experience level of sales person etc.). Either an offer was applied (sale) or it wasn't (no sale). However, the cases with the event of a sale amount to 500 and the one with no sale amount to 10k. Is that a problem? And if it is, how does site affect my estimates/odd ratios and what are possible solutions?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
bree
  • 91
  • 2
  • 7
  • 2
    I've added [tag:rare-events], which addresses this exact issue. You may benefit from reviewing posts bearing the tag. – Sycorax Feb 16 '18 at 16:51
  • I read that some people suggest bootstrapping, I applied that to my model, but the odds ratios didn't change too much. All that happened is that almost for all independent variables the confidence interval became smaller. In other posts, I read that for a logistic regression the imbalance of the response variable does not matter. Anyone who can provide more details on my particular case? – bree Feb 16 '18 at 18:57

0 Answers0