Detecting Heterogeneity within Demographic Predictors of Cognitive Test Scores
using Mixtures of Regressions 
   Lead Investigator:    William Goette
   Institution      :    University of Texas Southwestern Medical Center
   E-Mail           :    william.goette@UTSouthwestern.edu
   Proposal ID      :    1561



   Proposal Description:
   Identify whether within-group heterogeneity for predictors like race and         ethnicity can be accounted for via latent classes identified using finite         mixtures of regressions. The implication of these results would be whether       using group variables to predict test performance is valid for all individuals   as it is expected that, though an individual identifies as a certain race,       their experiences are unique from others in that same group. The question to be  answered in this case is whether those unique experiences convey unique           cognitive profiles.                                                               Aims:                                                                             Hypothesis - Latent classes of individuals will be found such that demographic   prediction of raw scores on cognitive tests is not sufficiently summarized by a  single regression but a unique regression for each class.                         Hypothesis - Within-group heterogeneity will be observed primarily for years of  education, race, ethnicity, and sex.                                             Hypothesis - This heterogeneity reflects differential levels of discrimination,  socioeconomic opportunity, and educational quality within individuals of the     same demographic traits.