Question on model weights

Avatar image
Aug 18, 2014
Lityx agent wrote
If you took all responders and 1/3 of non-responders, I would historically build models using the Weight statement since not including it would affect the model calibration. In addition, the performance statistics such as lift need to be off the weighted data. How would you recommend this be handled in LityxIQ?
Reply
1 Answer
Avatar image
Aug 18, 2014
Lityx agent wrote
The equivalent of using the weight statement in SAS, LityxIQ doesn’t do that. However, that is on purpose, and technically I would argue that weighting or not would have no or little effect on the estimated model coefficients if the dataset is of any reasonable size. This is because the multivariate distribution doesn’t shift due to the down-sampling of non-responders… we are just using fewer observations from that distribution. But using a weight statement could potentially affect standard errors of those estimates, probably producing smaller standard errors that you would get otherwise.