R/combine.mi.R

In vahidnassiri/mifa: Multiple Imputation for Exploratory Factor Analysis

# This function applies Rubin's rule to combine results from different imputations
combine.mi <- function(mi.parm.est,mi.parm.cov){
  # Input variables:
  # mi.parm: a matrix containing estimated parameters in each imputation
  # as its rows
  # mi.parm.cov: a list contaning the covariance matrix estimated within 
  # each imputation.
  M=length(mi.parm.cov)
  parm.est=apply(mi.parm.est,2,mean)
  est.diff=scale(mi.parm.est , center = TRUE, scale = FALSE)
  Sample.cov.parm1=NULL
  for (i in 1:M){
    Sample.cov.parm1[[i]]=(est.diff[i,])%*%t(est.diff[i,])
  }
  sample.cov.parm1=Reduce('+',Sample.cov.parm1)/ (M-1)
  Factor=(M+1)/M
  sample.cov.parm = Factor* sample.cov.parm1
  mean.cov.parm=Reduce(`+`, mi.parm.cov)/M
  parm.cov=mean.cov.parm+ sample.cov.parm
  return(list(parm.est=parm.est,parm.cov=parm.cov,between.cov=sample.cov.parm1))
}