2

I'm looking for a way to fit a conditional logit on grouped data in R. I have multinomial choice counts with response- and individual-specific variables. Unfortunately, the expanded dataset would be too huge to work with it. Thus I can't use most of the packages I know, mlogit for example. I discovered vglm from the VGAM package but it doesn't work if one explanatory variable is missing for a response category.

Has someone tried something similar in R ?

Thanks!

Cabana
  • 419
  • 2
  • 7
  • I've not used it myself so I'm not sure it will meet your specific requirements, but there is a package called `mclogit` that might do what you need. – David_B Jun 04 '16 at 11:33
  • Thanks for your answer! I've already tried `mclogit`. It only fits conditional models (with (mixed) or without random effect). So only response specific variables can be used. – Cabana Jun 04 '16 at 11:52

1 Answers1

1

You can use clogit in the survival package by setting the strata to your group id.

nikademus
  • 63
  • 6