0

I have data with some ordinal component of Injury severity, measured as property damage, Fatal or Killed.

Is there an R package for the Zero-inflated ordered Probit / Logit models?

Code newbie
  • 11
  • Welcome to the site. Could you say more about how damage is measured; particularly, why it is ordinal? It's not clear to me what your possible values are.

    Also, why do you think a regular ordinal logistic model would not be good?

    Usually, zero-inflation models are for count data, because the usual models for count data (Poisson and negative binomial) don't deal well with zero-inflated data. There's also zero-inflated normal, but that's less used.

    Why do you need "zero inflated ordinal logistic"?

     – Peter Flom Jul 02 '23 at 14:57
  • The response variable has 0, 1, 2,3 classified as no injury, minor, seriously injured or fatal. hence the need for zero-inflated model. – Code newbie Jul 02 '23 at 15:05
  • The data is also zero-inflated. They are a lot of zero in the model. Since it is an accident data, I can really find much on negative Binomial and Poission for thse types of problems – Code newbie Jul 02 '23 at 15:06
  • 2
    Nothing you've said so far shows that a zero-inflated model is needed. Nor is this count data, so negative binomial and Poisson are not right. You can use "regular" ordinal logistic. The fact that there are lots of 0s doesn't matter. – Peter Flom Jul 02 '23 at 15:26

0 Answers0