Longitudinal Measurement Invariance of Multigroup Ethnic Identity Measure (MEIM) among Ethnically/Racially Diverse Adolescents

• In Event: 3-066 - “What’s Measurement Got to Do with It?” Testing Ethnic-Racial Equivalence of Identity-Related Measures

Integrative Statement

Research on ethnic/racial identity (ERI) has grown considerably over the past few decades. In part, the growth in this literature is attributable to easily-administered measures that can be used across ethnic/racial and age groups. Perhaps the most popular of these measures is The Multigroup Ethnic Identity Measure (MEIM; Phinney, 1992). Despite the popularity of the measure, there is less work investigating its psychometric properties and measurement invariance. Since developmental scholars are often interested in investigating change in ERI over time, it is important to consider the consistency and reliably of the measure for understanding change in stability in ERI among diverse populations was over time.

Methods: Data were drawn from a large, 3-year longitudinal study of 405 diverse adolescents attending public high schools in New York City. Participants were 62.72% female, 34.32% Asian, 12.10% Black, 25.19% Hispanic/Latinx, 23.21% White, and 4.20% “other” race, and their average age was 15.23 (SD=0.67). Analyses were conducted in Mplus 7.4 and appropriate analysis options for our data, recommended indexes and cutoffs were applied to evaluate the model fit. Specifically, our analysis focused on the two MEIM subscales of identity exploration (6 items) and commitment (8 items) across 3 years.

Results: MEIM Exploration – the full model with all six items did not yield a good fit using maximum likelihood estimation with robust standard errors (CFI=0.838, TLI=0.810, RMSEA=0.058, 90% CI [0.049 0.066], SRMR=0.072). As a result, two items were removed (Table 1). Dropping these two items improved the model fit and the revised configural model had a good fit (CFI=0.926, TLI=0.904, RMSEA=0.047, 90% CI [0.033 0.061], SRMR=0.058). With this new model, configural, metric (-2∆LL(df=6)=3.58, p=0.64), and scalar (-2∆LL(df=7)=8.08, p=0.70) invariance with equal factor variance and means were established, indicating the factorial structure, factor loadings, and intercepts could be considered equal across the three time points (3 years).
MEIM Commitment – similarly, the full model with all eight items did not demonstrate a good fit (CFI=0.897, TLI=0.886, RMSEA=0.046, 90% CI [0.040 0.053], SRMR=0.07). Therefore, two items were dropped (Table 1), and the revised model had a better fit (CFI=0.927, TLI=0.916, RMSEA=0.046, 90% [0.037 0.055], SRMR=0.055). With the new model, configural, metric (-2∆LL(df=10)=11.96, p=0.36) and scalar (-2∆LL(df=9)=10.13, p=0.24) invariance with equal factor variance and means were established, indicating the factorial structure, factor loadings and intercepts could be considered equal across the three time points (3 years).

Taken together, the revised exploration and commitment subscales demonstrate equality and measurement equivalence in the MEIM scale across the first three years of high school in a diverse sample of adolescents. Implications and recommendations for dropping and retaining items along with the establishment of invariance over three years of high school among diverse adolescents will be discussed.

Authors

• Tiffany Yip, Fordham University
  Presenting Author

• Yuen Mi Cheon, Fordham University
  Non-Presenting Author