0

Multinomial regressions necessitate a reference level, and one can change the reference level using options like relevel(), but is there an alternative reference, that isn't one of the levels?

My datasets include DVs with 3-7 'levels', but they are not ordered in any kind of way, and there isn't a natural baseline, they are 3-7 separate outcomes that I am trying to characterize in terms of the IVs.

Progman
  • 16,827
  • 6
  • 33
  • 48
Maria Gold
  • 59
  • 3
  • 6
  • Did you already see https://stackoverflow.com/a/3874120/15685387 and the other non-`relevel`-related answers in this thread? You could also try reading the following UCLA document on this topic: https://stats.oarc.ucla.edu/stata/output/multinomial-logistic-regression/. It is written for stata but your question doesn't seem overly r-specific. – 542goweast May 24 '23 at 22:28
  • I did see that, and maybe I am not fully understanding, but it seems like this still compares all levels to one reference. My data does not have one 'natural' reference level – Maria Gold May 25 '23 at 23:07

0 Answers0