Tessa posted on Sunday, October 18, 2015 - 11:45 am
I plan to use MPlus multiple imputation to impute missing data on a continuous DV (measured across 5 time points) and several covariates, also with some missing data, prior to proceeding with GMM for continuous variables.
Several of these variables, including my DV, are scaled scores (the sum of several individual items). My question is whether it is best to impute the scaled scores or each of the individual items that make up these measures? If it is best to impute the individual items, how would one then compute scaled scores in MPlus upon completion of the multiple imputation procedure? I wish to use these scales as they are pre-existing standardized measures.
I am not sure we know the answer to that. But it seems best to work with the scale scores instead of the individual items because I assume the scale score is already taking into account item missingness, so you can focus on person missingness.
But always ask yourself if you couldn't do the analysis without imputation, instead going straight to ML or Bayes, including the covariates in the model. Imputation is typically used when the imputation analysis involves a different set of variables than the analysis of the model.
I plan to perform a test of common method bias. Specifically, I have two CFA models, one which has a method factor, and one that does not. I know that there is a way to impute composites for each scale based on the indicator loadings in the model with a method factor in Amos. Can this be done in Mplus as well? If yes, could you point me to an example? Your assistance will be much appreciated.