African-American Adolescent Male and Female Measurement Invariance within the Dual-Factor Model of Mental Health

Background/Purpose:  Promotion of the well-being of adolescents requires movement beyond the identification and treatment of psychological and behavioral problems. However, many services provided to adolescents with the intention of supporting their transition into adulthood are often guided by a deficiency focused mental health perspective that focuses on symptom reduction. Research on the dual-factor model of mental health (DFM) has provided evidence that attention to positive indicators of wellness (e.g., subjective well-being) is integral to the promotion of children’s and adolescents’ positive development (Suldo & Shaffer, 2008). Assessment under the DFM focuses on two dimensions:  positive (positive affect and subjective well-being) and negative indicators (negative affect) of mental health (Suldo et al., 2011). Support for this dual factor conceptualization and its predictive validity has some limitations. The purpose of this study was to evaluate the structural and predictive validity of the DFM.

Methods: Sample:Data from Wave 1 of the Add Health dataset was used in this study. The sample consisted of African-American male (n = 447) and female (n = 404) adolescents (ages 15 – 17).

Measures:The dual factor measure consisted of three subscales: life satisfaction, positive affect, and negative affect. Four items were selected from the Feelings Scale to operationalize life satisfaction. Items from the Feeling and Protective Factors Scales were used to create the Positive Affect Scale. Four items were selected from the Feelings Scale to assess the extent to which participants experience negative affect.

Analysis: The construct validity of the DFM measures was examined by exploring the factor structure of the hypothesized measure using Exploratory Structural Equation Modeling (ESEM). A series of Multigroup Confirmatory Factor Analysis (MG-CFA) models were used to evaluate the configural, metric and scalar measurement invariance of the DFM between African-American males and females.  Equality constraints on model parameters to the multigroup measurement model between the two groups and evaluating the impact of those added constraints on model fit was used to determine the presence of measurement invariance.

Results: The results only provide support for partial measurement invariance at the metric and scalar levels. At the metric level, two items from the Negative Affect subscale were found not to be equivalent between African-American males and females. The fit statistics for the metric model was CFI = .947, RMSEA 95% CI = .041(.034, .047) and the change in χ2 between the configural and metric models was not significantly different, Satorra-Bentler Scaled D χ2 (df) = 7.36(23), p>.05.  At the scalar level, 5 items taken from all three subscales failed to demonstrate equivalency between the two groups. The fit statistics for the model were CFI = .957, RMSEA = .028 95% CI (.020, .035). The Satorra-Bentler Scaled  D χ2 (df) = 6.22(10), p>05 comparing the metric to the scalar model.

Implications: The findings provide evidence that any comparison between African- American male and female adolescents using these sets of indicators of the DFM will need to take into account the presence of nonequivalence at the scalar level for items present on each of the subscales.