Poster #65 - A Bifactor Model of Internalizing Problems in Early Childhood: Relations with Future Peer Victimization

• In Event: 1-061 - Poster Session 02
  In Poster Session: PS 02 Section - Developmental Psychopathology

Integrative Statement

Prior research has found that there are reciprocal relations between peer victimization and internalizing problems. However, a majority of this research has been cross-sectional, defined social withdrawal and peer victimization as a unidimensional construct, and has taken place in later developmental periods. The first goal of the current study (N = 238, 56.3% boys, M age = 44.92 months, SD = 5.32) was to test a bifactor model of internalizing problems (i.e., anxiety and social withdrawal) which allows for a general anxious-avoidant withdrawal factor as well as a specific unsociability (i.e., withdrawal without anxiety) factor and a specific anxiety without withdrawal factor. The second goal of the study was to examine how the different factors predict changes in peer victimization over time.

Internalizing problems were measured at Time 1 (T1) in the spring using observer reports of the Anxious-fearful and Asocial subscales from the Child Behavior Scale (Ladd & Profilet, 1996). Physical and relational victimization were measured at T1 and Time 2 (T2) in the fall approximately six months after T1, using the Preschool Peer Victimization Measure- Teacher Report-Revised (Ostrov, 2008). These scales were reliable in the current sample (Cronbach’s α’s range from .66 - .93). Observer informants spent approximately nine hours a week for two months in the classroom. Models were tested using Structural Equation Modeling (SEM) techniques and the WLSMV estimator within Mplus (Muthén & Muthén, 1998-2017) because of the categorical nature of the indicators. To test model fit, the likelihood ratio χ2 was used in conjunction with the fit indices, root mean square error of approximation (RMSEA) and comparative fit index (CFI). RMSEA ≤ .08 and CFI > .95 indicated adequate fit. Chi-square difference tests were used to statistically compare the nested models. Alternative fit indices, such as Percentage Uncontaminated Correlation (PUC), Explained Common Variance (ECV), Omega H, and H were also examined.

The bifactor model provided an adequate fit to the data [χ2 (25) = 65.45, p < .001, CFI = 1.00, RMSEA = .08] and a significant improvement in model fit compared to the two-factor model [Δχ2 (9) = 61.50, p < .001]. There was evidence for multidimensionality of the internalizing construct with varying degrees of reliability across factors. The standardized factor loadings and alternative statistics are shown in Table 1. The predictive validity of the model was evaluated by testing pathways from the bifactor model to T2 relational and physical victimization while controlling for T1 relational and physical victimization, gender, and age. The predictive model provided an adequate fit to the data [χ2 (79) = 140.96, p < .001, CFI = .99, RMSEA = .05]. The standardized regression coefficients of the model are shown in Figure 1. Overall, results suggest that anxiety without withdrawal may lead to increased relational victimization, whereas anxious-avoidant withdrawal may lead to decreases in victimization. This decrease in relational victimization may reinforce withdrawal behavior. These findings demonstrate how different facets of internalizing problems relate to future peer victimization, which can help inform intervention efforts.

Authors

• Kristin Perry, University at Buffalo, SUNY
  Presenting Author

• Jamie Ostrov, University of Buffalo
  Non-Presenting Author