CORRECTING FOR MEASUREMENT ERROR IN LATENT VARIABLES USED AS PREDICTORS.

Area of Science:

• Educational Research
• Psychology
• Sociology

Background:

• Racial and gender disparities persist in Science, Technology, Engineering, and Mathematics (STEM) fields.
• Existing models may not fully account for measurement error in latent variables predicting STEM retention.
• Understanding factors influencing STEM persistence is crucial for developing effective interventions.

Purpose of the Study:

• Introduce a novel methodological approach, the Mixed Effects Structural Equations (MESE) model.
• Investigate the roles of pre-college mathematics proficiency and personality in STEM retention.
• Quantify the contribution of these factors to racial and gender disparities in STEM.

Main Methods:

• Developed the Mixed Effects Structural Equations (MESE) model, integrating structural equations modeling and item response theory.
• Utilized data from the 1997 National Longitudinal Survey of Youth (NLSY).
• Applied the MESE model to analyze predictors of STEM retention in higher education.

Main Results:

• Prior mathematics proficiency and personality traits were found to be significant, previously underestimated predictors of STEM retention.
• These factors collectively explain substantial portions of the observed racial and gender gaps in STEM persistence.
• The MESE model effectively addressed measurement error bias in generalized linear models.

Conclusions:

• Pre-college mathematics proficiency and personality are critical determinants of STEM retention.
• Interventions aimed at increasing STEM persistence should consider these factors, particularly for women and under-represented minorities.
• The MESE model offers a robust framework for analyzing complex relationships in educational research.