Attrition of participants is inevitable in longitudinal data collection. If participants who missed follow-up assessments are systematically different in the measured factors from the participants with full data, the assumptions that Mplus’ weighted least square estimator relies on are violated (Asparouhov and Muthén, 2010). To check the appropriateness of this assumption, we tested whether the factor models of alcohol use and antisocial behavior at each point are equivalent across individuals who did and did not miss subsequent measurements by conducting multiple-group CFA with individuals with or without the subsequent measurements as subgroups at each occasion.
