We used multiple imputation to account for missing responses, given that multiple imputation is more flexible to handle missing data for a mixture of categorical and continuous variables (Enders, 2010). Multiple imputation has also been shown to be robust and provide unbiased results even for high rates of missing data (Graham et al., 2003). Specifically, we used maximum likelihood estimation method to create ten imputed datasets in Mplus. The data imputation model included all of the observed study variables included in the SEM model. We conducted subsequent analyses with the imputed datasets and the final parameter estimates, standard errors, and goodness-of-fit statistics of the SEM model (representing the average results across 10 imputed datasets) were obtained with the automatic aggregation procedure implemented in Mplus (Rubin, 1987).
