Examining response patterns in attitudinal and personality scales using advanced psychometric modeling techniques

He, Siqi

Permalink

https://hdl.handle.net/2142/127259 Copy

Description

Title

Examining response patterns in attitudinal and personality scales using advanced psychometric modeling techniques

Author(s)

He, Siqi

Issue Date

2024-12-05

Director of Research (if dissertation) or Advisor (if thesis)

• Culpepper, Steven Andrew
• Douglas, Jeffrey A
• Kern, Justin

Doctoral Committee Chair(s)

Zhang, Susu

Committee Member(s)

Koehn, Hans Friedrich

Department of Study

Psychology

Discipline

Psychology

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Personality and Attitudinal Assessment
• Latent Variable Modeling
• IRTree
• Latent Class Model

Abstract

Chapter 1: In recent years, item response tree (IRTree) approaches have received increasing attention in the response style literature due to their capability to partial out response styles as separate latent variables. Furthermore, for non-cognitive psychological constructs, it has been argued that the underlying response process may be better described as following an ideal-point process rather than the dominance process often hypothesized to underlie cognitive constructs. In this study, we extended the family of IRTree models to handle ideal-point (i.e., unfolding) responses. An IRTree-based unfolding model is proposed to decompose the observed responses into two decision nodes, one for considering agreement with a hyperbolic cosine model (HCM) and another for considering the intensity of agreement (i.e., extreme response style trait) with a graded response model (GRM). We conduct a simulation study to evaluate the performance of parameter recovery. The model parameters are estimated in a full Bayesian approach with the JAGS (Just Another Gibbs Sampler) freeware. With multiple model comparisons, we demonstrate the advantages of the proposed model in two real data examples, one regarding attitudes toward capital punishment and another regarding attitudes toward censorship. The obtained item parameter estimates offer useful insight into the role extreme response style and content traits play within the IRTree framework. Chapter 2: Diagnostic models (DM) are widely utilized for binary attributes and response data, offering valuable insights in both exploratory and confirmatory analyses. The sparse latent class model (SLCM) has shown promise in such applications. However, in fields like educational and psychological measurement, ordinal data often provide richer information to capture latent structures, where binary attributes may fall short in describing complex response patterns. To address this gap, we extend the SLCM to accommodate ordinal attributes, enabling a more comprehensive exploration of the relationships between attributes and response patterns. We also establish strict and generic identifiability conditions for this extended model. To demonstrate its applicability, we apply the model to the Short Dark Triad dataset in an exploratory way to uncover the underlying personality structure. Comparisons reveal that the extended SLCM achieves better model fit than exploratory factor models for this dataset. Additionally, we propose a Gibbs sampling algorithm to efficiently recover the model parameters, confirmed through Monte Carlo simulation studies. This work expands the utility of diagnostic models by introducing ordinal attributes, broadening their applicability across various domains. Chapter 3: Diagnostic models (DM) have been widely employed to classify respondents' latent attributes in cognitive and non-cognitive assessments. The integration of response times (RT) with DM presents additional evidence to understand respondents' problem-solving behaviors. While recent research has explored using sparse latent class models (SLCM) to infer the latent structure of items based on item responses, the incorporation of RT data within these models remains underexplored. This study extends the SLCM framework to include RT, relaxing the conditional independence assumption between RT and latent attributes given individual speed. This adaptation provides a more flexible framework for jointly modeling RT and item responses. The method is applied to data from the Fisher Temperament Inventory, providing findings that offer a novel perspective on using DM with RT in personality assessments. Additionally, a Gibbs sampling algorithm, enhanced with variable selection techniques, is proposed for parameter estimation. Results from Monte Carlo simulations demonstrate the algorithm’s accuracy and efficiency.

Graduation Semester

2024-12

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/127259

Copyright and License Information

2024 by Siqi He. All rights reserved.

Owning Collections