0

I recently fitted a Graded Response Model to my data using R's latent trait modelling package. I tried to add my fitted Graded response model in mice package to impute missing data but i am failing to access the mice algorithm to edit. My question is does it make any difference if i use the polr() function in mice or if i use my graded response model that i fitted using latent trait modelling package to impute data since both these functions are derived from polr() function in R's mass package

tonderai
  • 1
  • 3
  • What does your GRM look like? – SimonG Oct 05 '14 at 14:34
  • To be honest, I couldn't find how exactly the `ltm` package makes use of `polr`. The documentation merely states `MASS` as a dependency and says that some code is based on the `polr` function. There is no call to `polr` within `grm` or the functions called in the process. The "polr" method in `mice` on the other hand actually does call `polr`. Being not familiar with the `ltm` package and its code, I find this hard to tell. – SimonG Oct 05 '14 at 15:26
  • Thank you very much SimonG. I suppose then my only option is to add my fitted grm model in mice which is what i am finding challenging – tonderai Oct 06 '14 at 09:22
  • @tonderai please don't reply with code segments in comments, [edit] the question and update it there... – T J Oct 12 '14 at 08:06
  • I am not familiar with Graded Response Models. Is that a deterministic or stochastic function? If it is deterministic, you could use passive imputation. – Paul de Barros Apr 14 '16 at 23:37
  • It's not deterministic. It's based on an exponential probability function see Graded Response Model by Samejima (1969) for example – tonderai Apr 16 '16 at 02:45

0 Answers0