I have a longitudinal cohort which has 2 treatment groups and 4 time points. Let's say the model is $Y = Group + Time + Group*Time + Age + Gender + Race + Household Size$. Here are my analytic steps:
1. There are missing in the outcome and covariates so I imputed 20 times using mice.
2. I fitted linear mixed effects model using lme4 on each set of imputation, then pooled them into one.
3. I wanted know the mean and CI for group 1 at time 1, group 1 at time 2, etc. I was told to treat the pooled LME model as my new model to use linear combination of the coefficients. However, the object I got from pool has a class of mipo. It is not a model object and does not work with glht. Can I make up my own model object to feed in glht? Is there a better way to do it?



Browse other questions tagged or ask your own question.