Education and Human Sciences, College of (CEHS)

 

First Advisor

James Bovaird

Date of this Version

5-2024

Document Type

Thesis

Citation

A thesis presented to the faculty of the Graduate College at the University of Nebraska in partial fulfillment of requirements for the degree of Master of Arts

Major: Educational Psychology

Under the supervision of Professor James Bovaird

Lincoln, Nebraska, May 2024

Comments

Copyright 2024, Sarah Hammami. Used by permission

Abstract

Regression discontinuity designs are valuable quasi-experimental designs, especially when randomized control/clinical trials are ethically or practically challenging. Latent variables are prevalent in the social, behavioral, and educational (SBE) sciences—measurement assumptions have to be tested and met when using such variables. Not meeting those assumptions leads to biased results and (likely) erroneous inferential making. One way to minimize the impact of measurement assumption violation is to use latent variable modeling such as structural equation modeling (SEM). Soland, Johnson and Talbert (2022) conducted two simulation studies testing the effect of using a sum score (when the assumptions for using such a score are not met) versus using a latent variable model to model the outcome on the bias and Type II error rates of the treatment effect. For the two simulation studies, bias and Type II error rates were optimal when the outcome was modeled within an SEM framework. Extending the work of Soland et al. (2022), there is evidence that measurement error, number of items, and sample size have an effect on SEM outcomes. Findings suggest that having numerous indicators with high measurement accuracy and a large sample size (> 2,000) leads to the most optimal treatment effect. That being said when one has a low level of either the number of indicators, their measurement accuracy, or sample size, increasing at least one of them improves the accuracy and power of the treatment effect.

Advisor: James Bovaird

Download

Share

COinS