Measurement Equivalence of Parenting Scales for Hispanic and Non-Hispanic White Parents

Introduction: In order to validly compare if an intervention effect differs across ethnicity, the scales that measure the outcomes need to have an equivalent factor structure across the ethnic groups being compared. To determine equivalence, four sequential invariance models are typically analyzed: configural, metric, scalar, and strict. This study reports confirmatory factory analyses and construct invariance for several parenting and family interaction scales among non-Hispanic White and Hispanic parents and children from divorced/separated families. We also discuss special concerns that need to be addressed when analyzing measurement invariance using categorical items rather than continuous items.

Methods:To ensure that the differences of intervention effects were not attributed to incompatibility of the measures, we analyzed measurement equivalence models comparing 493 non-Hispanic White families and 261 Hispanic families using data from the New Beginnings Program multi-site effectiveness trial. Parent ages ranged from 19 to 64 and children’s ages ranged from 3 to 18. We analyzed 19 parenting and family interaction scales and subscales including the Children’s Report of Parental Behavior Inventory (CRPBI; Schaefer, 1965), Parent-Adolescent Communication Scale (Barnes & Olson, 1982), Children’s Perception of Interparental Conflict (Grych, Seid, & Fincham, 1992), and Oregon Discipline Scale (Oregon Social Learning Center, 1991).

Results: For all scales and subscales, we sequentially tested the configural invariance model, which constrains the factor structure to be equivalent, the metric invariance model, which constrains the loadings to be equivalent, the scalar invariance model, which constrains the thresholds to be equivalent, and the strict invariance model, which constrains the unique variances to be equivalent. Some of the scales had items that were measured on ordered categorical scales whereas others had dichotomous items. We used recommendations of the skewness and kurtosis by Rhemtulla, Brosseau-Liard, and Savalei (2012) to determine if we could treat the categorical items as continuous variables. The CRPBI exhibited convergence issues, but utilizing a different set of identification constraints allowed for metric and scalar invariance testing (Yoon & Millsap, 2007). Fifteen out of 19 scales and subscales exhibited scalar invariance.

Conclusions: The findings demonstrated that any intervention effect differences between Hispanics and non-Hispanic Whites can be validly compared on the majority of the scales because the measurement structure is equivalent across the two groups. Establishing measurement equivalence is important in determining if there are mental health disparities and differential intervention effects.