This study investigated the sensitivity of common fit indices (i.e., RMSEA, CFI, TLI, SRMR-W, and SRMR-B) for detecting misspecified multilevel SEMs. The design factors for the Monte Carlo study were numbers of groups in between-group models (100, 150, and 300), group size (10, 20, 30, and 60), intra-class correlation (low, medium, and high), and the types of model misspecification (Simple and Complex). The simulation results showed that CFI, TLI, and RMSEA could only identify the misspecification in the within-group model. Additionally, CFI, TLI, and RMSEA were more sensitive to misspecification in pattern coefficients while SRMR-W was more sensitive to misspecification in factor covariance. Moreover, TLI outperformed both CFI and RMSEA in terms of the hit rates of detecting the within-group misspecification in factor covariance. On the other hand, SRMR-B was the only fit index sensitive to misspecification in the between-group model and more sensitive to misspecification in factor covariance than misspecification in pattern coefficients. Finally, we found that the influence of ICC on the performance of targeted fit indices was trivial.

No abstract

Disciplinary exclusion practices are on the rise nationally, as are concerns about their disproportionate use and lack of effectiveness. This study used data from the Special Education Elementary Longitudinal Study to examine patterns and predictors of disciplinary exclusion over time. Students with emotional/behavioral disorders were most likely to be excluded and be excluded multiple times, followed by students with attention deficit-hyperactivity disorder and students with learning disabilities. For all student groups, being excluded in the first wave was a strong predictor of being excluded at later points in time. Student gender (male students) and ethnicity (African American students) were associated with a greater probability of exclusion over time. Students with higher social skills, as reported by teachers, had a lower probability of being excluded over time. Implications for practice, policy, and future research are discussed.

Several researchers have recommended that level-specific fit indices should be applied to detect the lack of model fit at any level in multilevel structural equation models. Although we concur with their view, we note that these studies did not sufficiently consider the impact of intraclass correlation (ICC) on the performance of level-specific fit indices. Our study proposed to fill this gap in the methodological literature. A Monte Carlo study was conducted to investigate the performance of (a) level-specific fit indices derived by a partially saturated model method (e.g., [Formula: see text] and [Formula: see text]) and (b) [Formula: see text] and [Formula: see text] in terms of their performance in multilevel structural equation models across varying ICCs. The design factors included intraclass correlation (ICC: ICC1 = 0.091 to ICC6 = 0.500), numbers of groups in between-level models (NG: 50, 100, 200, and 1,000), group size (GS: 30, 50, and 100), and type of misspecification (no misspecification, between-level misspecification, and within-level misspecification). Our simulation findings raise a concern regarding the performance of between-level-specific partial saturated fit indices in low ICC conditions: the performances of both [Formula: see text] and [Formula: see text] were more influenced by ICC compared with [Formula: see text] and . However, when traditional cutoff values (≤ 0.06; ≥ 0.95;≤ 0.08) were applied, [Formula: see text] and [Formula: see text] were still able to detect misspecified between-level models even when ICC was as low as 0.091 (ICC1). On the other hand, both [Formula: see text] and [Formula: see text] were not recommended under low ICC conditions.

This study investigated the construct validity of measures of teacher-student support in a sample of 709 ethnically diverse second and third grade academically at-risk students. Confirmatory factor analysis investigated the convergent and discriminant validities of teacher, child, and peer reports of teacher-student support and child conduct problems. Results supported the convergent and discriminant validity of scores on the measures. Peer reports accounted for the largest proportion of trait variance and non-significant method variance. Child reports accounted for the smallest proportion of trait variance and the largest method variance. A model with two latent factors provided a better fit to the data than a model with one factor, providing further evidence of the discriminant validity of measures of teacher-student support. Implications for research, policy, and practice are discussed.