Weighted least squares estimation for ordinal outcome variables in Mplus was used (Muthén and Muthén, 2007). This estimator allows missing data as a function of the observed covariates. Specifically, data are assumed to be missing at random with respect to the outcome variables but missingness is allowed to be dependent on the covariates. When there are no covariates in the model, this is analogous to pairwise present analysis in which all available data are used to estimate each pairwise association. The weighted least squares estimation has been shown to be consistent and more efficient than estimators based on listwise deletion (Asparouhov and Muthén, 2010). Chi-square tests of model fit were computed taking into account the non-independence of observations due to cluster sampling. Other indicators of model fit were the Root Mean Square Error of Approximation (RMSEA), for which values below .05 are generally taken to indicate good fit, and the Comparative Fit Index (CFI), for which values greater than .90 are taken to indicate reasonably good fit to data (Kline, 2005).
