Evaluation of a comprehensive multidimensional model of bipolar spectrum psychopathology through statistical and machine learning methods

Chia, Talia Rebecca Berson

Permalink

https://hdl.handle.net/2142/125517 Copy

Description

Title

Evaluation of a comprehensive multidimensional model of bipolar spectrum psychopathology through statistical and machine learning methods

Author(s)

Chia, Talia Rebecca Berson

Issue Date

2024-06-13

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

Kwapil, Thomas R

Doctoral Committee Chair(s)

Kwapil, Thomas R

Committee Member(s)

• Hankin, Benjamin L
• Briley, D. Ava
• Tang, Yan
• Xia, Yan

Department of Study

Psychology

Discipline

Psychology

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• bipolar
• multidimensional
• statistical modeling
• regression
• machine learning

Abstract

Researchers have proposed that bipolar disorders are better understood as a continuum, ranging from subclinical through clinical severity and impairment. Furthermore, there is compelling evidence to reconceptualize bipolar spectrum psychopathology as multidimensional. However, researchers have struggled to capture a consistent multidimensional structure. Based on an extensive literature review, a nine-dimensional model was proposed consisting of elation-euphoria, irritability, mood lability, energy-activation, impulsivity-disrupted reward sensitivity, unstable self-esteem, unstable sociability, decreased need for sleep, and lack of insight. This study evaluated the proposed multidimensional model in a multisite, non-clinically ascertained sample of young adults (n = 1264). Participants completed self-report assessments evaluating each domain, as well as a questionnaire assessing bipolar spectrum psychopathology broadly. The proposed model was evaluated using both traditional statistical techniques, such as confirmatory factor analysis and multiple linear regression, as well as machine learning techniques, such as parallel coordinates, elastic net regression, support vector regression, and artificial neural networks. Confirmatory factor analysis indicated that this model did not result in overall acceptable fit. However, both confirmatory factor analysis and regression analyses suggested that many dimensions represented relevant features of bipolar spectrum psychopathology. Both statistical and machine learning models indicated that this model accounted for approximately 60% of the variance of the HPS. Furthermore, statistical and machine learning techniques yielded comparable results. Therefore, next steps were proposed for refinement and evaluation of a comprehensive and theoretically valuable multidimensional model of bipolar spectrum psychopathology.

Graduation Semester

2024-08

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/125517

Copyright and License Information

Copyright 2024 Talia Chia

Owning Collections