0

I'm working with a client that wants cost differences between two treatment groups. They have little background knowledge but have seen in the literature folks doing GLM gamma/log-link, etc. and are suggesting we do the same. However, for their purposes and the ease of the analysis, I want to suggest using IPTW to balance between the groups and compare cost directly using those weights. I need to strengthen my argument and I'd be happy to take suggestions on where to go.

j_bro_rx
  • 63
  • 6
  • This isn't an either or - there's no conflict between the methods. You can weight GLMs just as you can weight OLS. – Affine Jun 12 '15 at 14:08
  • I agree. The argument is using ONLY GLM or ONLY IPTW. I would be open to GLM after IPTW but I'm not willing to only use GLM without some form of propensity scoring method. – j_bro_rx Jun 12 '15 at 17:13
  • That's an unusual stance to take. The vast majority of GLMs are done without propensity weights. – Jeremy Miles Jun 12 '15 at 17:28

0 Answers0