Multilevel imputation with parlmice

[bbachilles]

Hello and thanks so much for all the work you do on mice. Following the setup from the multilevel mice vignette, I was able to get single-core mice to run on a 15 column, 28K row dataset. It took about 30 minutes to run.

I have tried to replicate the process but instead use parlmice for the parallel mice implementation. However, I am getting an error: Error in checkForRemoteErrors(val) : 5 nodes produced errors; first error: invalid 'y' type in 'x && y'

The coding leading up to this is below. Is this a unique problem to me or is it that parlmice is not setup for multilevel models? Any help would be greatly appreciated?

# Create a new cluster ID as a placeholder
m_34_red <- m_34_red %>% group_by(NCESID) %>% mutate(NCESID = cur_group_id())

#First, run a 0 iteration model to see how mice sets up the data           
ini <- parlmice(m_34_red, maxit = 0)

#Set ID as class variable for 2l.norm
pred1 <- ini$predictorMatrix
pred1[,"NCESID"] <- -2 

#Declaring method as 2l.pan
meth1<-ini$method
meth1[which(meth1 == "pmm")] <- "2l.pan"

#Parallel 
seqTime_mlm_par <- system.time(imp_mlm <- parlmice(m_34_red[,c(3:21)], cluster.seed = 22901, 
                                           method=meth1, 
                                           predictorMatrix = pred1, print = FALSE))

[hanneoberman]

Hi @bbachilles, could you please share a reprex for this issue?

[stefvanbuuren]

We are going to retire parlmice() in favour of the much better behaved futuremice() available in mice 3.14.12. Hope that solves the problem.