A Comparison and Analysis of the Relevant Features in Regression Modeling with Latent Variables using Multi-Item Measures for Developmental Science

Authors

Christopher Clifford, Florida International UniversityFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Psychology

First Advisor's Name

Leslie D. Frazier

First Advisor's Committee Title

Co-Committee Chair

Second Advisor's Name

Stefany Coxe

Second Advisor's Committee Title

Co-Committee Chair

Third Advisor's Name

Timothy Hayes

Third Advisor's Committee Title

Co-Committee Chair

Fourth Advisor's Name

Eliza Nelson

Fourth Advisor's Committee Title

Committee Member

Fifth Advisor's Name

Nicole Fava

Fifth Advisor's Committee Title

Committee Member

Keywords

play, playfulness, stress, coping, regression, latent variable, structural equation modeling, sum score, factor score regression, structural after measurement

Date of Defense

6-16-2023

Abstract

In developmental science, a common and essential goal of research is to understand the relations that exist between constructs of interest. These constructs are not always directly observable and, in some cases, refer to abstract or theoretical factors. In those instances, the constructs end up being inferred from the results of measures or self-reported assessments taken across one or more items. The goal of this dissertation was to demonstrate the use of regression modeling methods for analyzing the relation between latent variables composed of multi-item measures.

This dissertation examined regression methods for multi-item measures in three separate studies. The first was an empirical study using a latent variable modeling framework to examine the roles of playfulness, stress, and coping. The second study used the data from the first study to compare four different approaches for estimating regression relationships with multi-item measures. The third study used a statistical simulation to demonstrate how data characteristics (i.e., sample size, effect size, and reliability) differentially impact the accuracy of different methods. Together, this dissertation examined the practical and theoretical considerations in regression analysis using latent variables. The findings from this dissertation have implications for developmental scientists and provide empirically grounded guidance for the selection of regression estimation methods to use for assessing latent variable relations.

Identifier

FIDC011155

Previously Published In

Clifford, C., Paulk, E., Lin, Q., Cadwallader, J., Lubbers, K., & Frazier, L. D. (2022). Relationships among adult playfulness, stress, and coping during the COVID-19 pandemic. Current Psychology, 1-10.

Recommended Citation

Clifford, Christopher, "A Comparison and Analysis of the Relevant Features in Regression Modeling with Latent Variables using Multi-Item Measures for Developmental Science" (2023). FIU Electronic Theses and Dissertations. 5420.
https://digitalcommons.fiu.edu/etd/5420