Bridging Disciplinary Boundaries (January 11 - 14, 2007)

Purpose: This study used a special type of structural equation model, the Multiple Indicators/Multiple Causes (MIMIC) model to incorporate the effects of differential item functioning (DIF) into the measurement of parent's beliefs regarding the importance of various academic and social skills for children entering kindergarten. Specifically this study examined the DIF of a measure of readiness beliefs between African American and European-American parents when social economic status was controlled.

Background: The readiness of children to enter elementary school prepared to successfully engage in the educational process continues to be a central challenge for public educations. Social workers have found themselves increasingly being called upon to engage parents in the process of addressing children's readiness needs. The study of how parent's school readiness beliefs serve as both a facilitator and barrier to children's readiness and parental involvement will be hamper by measures that are non-equivalent across groups. Measures of parent's readiness beliefs that are non-equivalent across groups can reduce the validity of research done using those measures. MIMIC models provide a very flexible and efficient way to study the DIF of measures. One research question guided this study: When social economic status was controlled for, was there any DIF of the measures between African and European American parents?

Methods: Using data from the Early Childhood Longitudinal Study-Kindergarten Class of 1998-99 (ECLS-K) the DIF of a readiness belief index was examined. Parent's readiness beliefs were hypothesized to consist of two latent variables: readiness beliefs related to children's social interaction and readiness beliefs associated with pre-school academic competencies. There were eight binary indicators for the academic readiness and four binary indicators for social interaction readiness. The two covariates in the model were social economic status and ethnicity. The models were estimated using Mplus 4.0. Mplus 4.0 allowed for the model to be estimated using weighted least squares mean and variance adjusted estimator (WLSMV). WLSMV efficiently addresses both the complex sampling design of the ECLS-K study and the binary nature of the indicators for the readiness latent variable.

Results: The fit indices for the MIMIC model indicated a good model fit (CFI=.97; TLI= .98 and RMSEA = .016). For the two covariates only the path between ethnicity and the indicators for academic readiness was significant, the unstandardized beta = -0.378, SE =.066, p<.01. The significance of path for the ethnicity but not social economic status indicates the DIF of the academic readiness factor between African Americans when compared to European American.

Implications: The validity of our measurement tools in social science is critical when we are trying to understand group differences. Understanding the DIF of our measurement tools will improve the validity of those tools and the validity of our research findings. This study also provides evidence of the association between parent's ethnicity and DIF of a readiness belief index. These finding, provide evidence that differences in readiness belief that may exist between African and European American parents is, in part, due to the DIF of the measures.