Appears Happy When Playing With Others: Measurement Models and Measurement Invariance of the Devereaux Early Childhoodhood Assessment

• In Event: Poster Session 06
  In Poster Session: PS 06 Topic - School Readiness/Childcare

Abstract

Accurately measuring the development of children’s social and emotional skills can be difficult (Diamond et al., 2013). Measurements must be able to capture developmental gradients in children’s behavior, accurately reflect contexts in which behavior occurs, and be reliable indicators of what they measure (Snow, 2011). Additionally, as classrooms and children become more diverse, we must ensure that the measures we use are ecologically valid and true across populations (Raver et al., 2007). One widely used measure of the social-emotional behavior of children aged three to five years is the Devereaux Early Childhood Assessment (Deca-P2; LeBuffe & Naglieri, 2012). Teachers rate the frequency of various behavior on a 1-5 scale to measure children’s positive social-emotional functioning (i.e. protective factors) and to screen children at risk for behavioral problems. The protective factors scale contains three subscales or latent factors, attachment, self-regulation, and initiative. While many studies using the Deca-P2 protective factors scale have supported the original 3 factor solution, some studies have also found evidence of individual items not loading onto the original factors or loading onto multiple factors. Moreover, limited empirical data has been reported on the measurement equivalence of the Deca-P2 across various subgroups (Ogg et al., 2010). Therefore, this study examined the factor structure of the Deca-P2 and attempted to determine if the structure is equivalent across child characteristics such as gender, race/ethnicity, and home language.
The data from this study came from a program evaluation research partnership between a local Head Start program and a research-based university. As part of this partnership, data on children’s cognitive, social/emotional, and academic functioning has been collected in the fall and spring from randomly selected Head Start classrooms. For this study, three and a half years of the Deca-P2 were compiled for analysis (n=2497), with some participants having data collection across the fall and spring and over multiple years. The final sample includes 1,210 children (male=658) with a mean age of 3.25 (SD=.69). The sample is ethnically diverse (e.g., White=173, Black=360, Hispanic=463, American Indian=41). The majority of this sample’s (n=463) home language was Spanish.
All the data for this study has been collected and complied. Analysis will be conducted using SPSS and Mplus. First, confirmatory factor analysis using structural equation modeling will be used to attempt to replicate the original factor structure of the Deca-P2. After the factor structure has been replicated, measurement equivalence tests will be used to explore whether the factor structure is similar across multiple subgroups including race/ethnicity, English language proficiency, and gender. Results from this study will help validate the underlying factor structure of the Deca-P2 with a more ethnically and linguistically diverse sample. Examining the measurement invariance will help researchers and teachers understand how scales and items function and whether teachers rate subgroups of children similarly. This has practical implications when considering whether a measure should be used to screen children: knowing whether a child’s score truly indicates a behavioral concern or is an artifact of bias is critical in helping teachers and families support children’s social-emotional development.

Authors

• Amber Beisly, University of Oklahoma
  Presenting Author

• Shinyoung Jeon, University of Oklahoma
  Non-Presenting Author

• Sherri L Castle, University of Oklahoma
  Non-Presenting Author