Hello all, I used the MICE procedure (of package "mice") to impute a dataset (I got m imputed datasets). Now I would like to fit a GLM with a poisson error distribution to regress a count variable on 14 continuous predictor variables and test for the significance of the different predictors by means of likelihood ratio tests (LRT).
I can estimate the pooled estimates and respective standard errors (of the m fitted GLMs) using function "pool" (which follows Rubin's 1987 rules). However, as far as I understand, the function "pool.compare" only allows to perform LRT on logistic models (presumably model with a binomial error distribution). Is this correct, or can I use this function to perform the a LRT for poisson models? Otherwise, would it be correct to average the residual deviance across the m fits for both the full and nested (in which the variable for testing is removed) models to posteriorly perform the chi-square test (difference of averaged deviance, difference of df)? Any help would be greatly appreciated. Duarte ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
