“…The FIML procedure is recommended in several studies on missing data in longitudinal data analysis because of unbiased parameter estimation, less convergence failures, high parameter and model fit estimation efficiency in the case of time-specific dropouts. This even holds for missing data proportions of 25 to 50% (Enders & Bandalos, 2001;Jeličić, Phelps, & Lerner, 2009;Raykov, 2005;Shin, Davison, & Long, 2009).…”