Hi everybody, I have a very practical question for the list. My colleagues and I are analyzing longitudinal repeated measures data in which the final dependent variable in our analysis is dichotomous (i.e., presence or absence of a substance use disorder). The remaining exogenous predictors and endogenous mediators are continuous and normally distributed. We are just now becoming familiar with the recent advances in multiple imputation and are wondering what procedures would be appropriate for the type of data described above. For example, I have heard it said that it may be appropriate to use a program like NORM for such data at the imputation stage, so long as the data sets are analyzed with a proper recognition of their properties (we plan to conduct structural equation modeling with manifest variables using Mplus and a weighted least squares estimator). Any suggestions would be much appreciated. If what I have described is unclear or more detail is needed, I would be more than happy to elaborate. Thank you very much. Alex Mason University of Alabama at Birmingham
