A Demonstration of the Utility of Latent Profile Analysis for Critical Quantitative Frameworks

• In Event: Interrogating Mixture Models in Pursuit of True Population Heterogeneity

Abstract

Critical quantitative approaches have called for researchers to engage in race-conscious and socially-just research when employing statistical methods (Garcia et al., 2018, p. 4; Gillborn et al., 2018). While not one method could help respond to this call, mixture modeling can assist researchers in advancing anti-racist and equitable research (Suzuki et al., 2021). Mixture modeling can help elucidate the nuances of black individuals and people of color (BIPOC), who are often depicted as homogenous groups (Neblett et al., 2016). Specifically, mixture modeling can facilitate person-centered research by explaining minoritized population heterogeneity through identified latent subgroups (Lanza & Cooper, 2016).
To assist researchers in developing anti-racist scholarship with mixture models, Suzuki and colleagues (2021) identified and outlined three moments of intervention in research that investigators can integrate critical quantitative principles, including QuantCrit tenets. The first moment to incorporate critical principles occurs early in the research process, during which the researcher actively develops and identifies racially just research question(s) and variable(s). In the second moment, researchers make critical and racially responsible decisions about the role of race. The last opportunity is especially important when using mixture modeling. During this moment, the results from a mixture model are suggested to be representations of data and not permanent classifications.
In the present study, we use latent profile analysis (LPA), with a QuantCrit framework and highlight the three moments for intervention. Specifically, this demonstration builds upon and expands on Young and Cunningham's (2021) study, which investigated the heterogeneity in math disposition among young Black females. In addition to Black girls, we explore Hispanic girls' math dispositions. Often when females of color participate in STEM fields, they experience race and gender discrimination in education and work environments (Johnson, 2011). Despite challenges, Black and Hispanic women have successfully attained positions in those fields (Carlone & Johnson, 2007). Therefore, it is crucial to understand and identify the factors that help increase Black and Brown females' representation and retention in STEM majors and careers.
The present study used the High School Longitudinal Study of 2009 (HSLS:09) to examine Black and Hispanic female students' math dispositions. LPA was employed in MPlus 8.7 (Muthén & Muthén, 1998). Profile enumeration was evaluated by estimating four model structures for 1- through 6- profile solutions using maximum likelihood estimation with robust standard errors (MLR). To determine the optimal number of profiles, approximate and relative fit indices were evaluated as Nylund-Gibson, and Choi (2018) recommend. The Block, Croon-Hagenaars (BCH) method was used to relate covariates and outcomes to profiles.
This study identified three distinct groups of Black and Hispanic female high school students with varied math dispositions. STEM academic major consideration, job attainment, and educational attainment varied across Black and Hispanic female adolescent profiles. Findings highlight the potential of LPA in developing and participating in racially- and socially-just research. LPA can distinguish and describe subgroups in diverse groups assumed to be homogeneous.

Authors

• Netasha Pizano, University of California - Santa Barbara
  Presenting Author

• Karen L. Nylund-Gibson, University of California - Santa Barbara
  Presenting Author

• Marsha M. Ing, University of California - Riverside
  Presenting Author