Problem passing family argument to fit.mult.impute for multiple imputation with modified Poisson using Hmisc package

177 Views Asked by At

I was previously able to run a modified Poisson regression using data created through multiple imputation with the Hmisc package. I ran the code a month ago without issue, but I'm now having problems using the same code and data. Specifically, I am getting an error for unused arguments for the family argument.

Here is a minimal version of the code I am using:

d <-datadist(data)
options(datadist = "d")

imp <- aregImpute (~ y + z, nk = c(o,3), tlinear = false, data, B=10, n.impute=10)

fit.mult.impute (x ~ y + z, glm, family = poisson(link = "log"), xtrans = imp, data, fitargs = list(x=true, y=true)

I get

Error in fit.mult.impute(x ~ y + z : unused argument (family = poisson(link = "log"))

I tried this with family = "binomial" which is used in several examples I found, and this produced the same error. I also tried family = stats::poisson and family = "poisson" which were in a similar post. Only if I remove the family argument completely will the code run, but without the correct model type.

Is there a new argument name or some other change to the package that I can't find which has caused my code to no longer work?

0

There are 0 best solutions below