1

Lets assume we have a dependent varible which can take on three values: 1, 2 and 3.

Is there any differences in running multiple binary logit regressions(ie. 1 vs 2 and 2 vs 3) or the multinomial logit regressions with 2 as the base group? Will the results be the same as the base group is the same in both cases? I've tried to look it up but can't seem to find an answer.

Thanks in advantage, have a nice weekend.

Regards, Martin

Frank Harrell
  • 74,029
  • 5
  • 148
  • 322
Martin
  • 11
  • 1

1 Answers1

1

The regression coefficients are the same, but I believe the standard errors are not. Using the full polytomous (multinomial) logistic model better exposes the true (large) number of parameters being estimated.

Frank Harrell
  • 74,029
  • 5
  • 148
  • 322