May 9, 2021

Multiple Imputation Madness

tl;dr summary: If you use multiple imputation in Mplus software, you can’t request a single pooled asymptotic covariance matrix of the parameter estimates (at least not as of version 8.5). The output file containing separate covariance matrices for each iteration can’t be easily read into other software. Here I show how to read the file into R and how to get a pooled matrix, for later use in getting standard errors of point estimates and making plots and things.