-1

I wanna solve my problem using multinomial logistic regression, mlogit R-package.

when I run the below code,

ml.data <- mlogit.data(dat, shape = "wide", choice = "Resp")
mlogit(Resp ~ A + B+ C+ D, ml.data)

I am getting this error: "system is computationally singular". I saw this post, but it does not help. I appreciated if anyone could help me.

user4704857
  • 469
  • 4
  • 18
  • 1
    Check out this post for an idea of the problem: http://stats.stackexchange.com/questions/32585/singularity-issues-in-multinomial-model-using-r – Vedda Dec 04 '15 at 22:13

1 Answers1

4

mlogit doesn't follow the same model command-line formulas as other models.

Try

mlogit(Resp ~ 1|A+B+C+D , ml.data)

and compare to

multinom(Resp ~ A+B+C+D , dat)

in package nnet. Also, check out ?mFormula and this answer

Community
  • 1
  • 1
David
  • 572
  • 3
  • 12