Improving Our Classification System for the Treatment of Individuals Who Have Experienced Traumatic Events: The Contribution of Unsupervised Statistical Learning to Our Existing Methods

Date

2016-05

Contributor

Advisor

Department

Instructor

Depositor

Speaker

Researcher

Consultant

Interviewer

Narrator

Transcriber

Annotator

Journal Title

Journal ISSN

Volume Title

Publisher

[Honolulu] : [University of Hawaii at Manoa], [May 2016]

Volume

Number/Issue

Starting Page

Ending Page

Alternative Title

Abstract

Rationally derived theories have had a limiting effect on the advancement of psychology as a science, compared to theories born out of or tested by empirical studies. As an example, while the diagnostic system (DSM) has been informed by science, the categories have not often been empirically derived (DSM-I, 1953; DSM-II, 1968; DSM-III, 1980, DSM-IV-TR, 2000; DSM-5, 2013). There is an emerging inclusion of empirical methods in the diagnostic classification system, as seen with some diagnostic categories of the DSM-5 (2013; Krueger, Derringer, Markon, Watson, & Skodol, 2012); however, there are many criteria and categories that have gone untested (Kramer et al., 2016). And, simply using hypothesis testing may not be sufficient in generating new knowledge. To improve our methods, we add to our current research and statistical methods through the use of unsupervised statistical learning, where data are allowed to tell their own story. Two statistical learning techniques, k-means cluster analysis and finite mixture modeling (Duda, Hart, & Stork, 2012; Hastie et al., 2009; James, Witten, Hastie, & Tibshirani, 2013) were applied to a data set collected on university students who had been displaced in the aftermath of Hurricane Katrina to understand the relationship between resource loss and stress. These techniques were used to demonstrate how to explore the data so that unanticipated knowledge could be distilled from the data. Findings showed that this data set was not easily studied using k-means cluster analysis, because the structure of the multivariate data did not contain clearly defined subgroups. Exploring the data using finite mixture modeling did, however, yielded possible areas of further research, such as the relationship of gains and losses to items on the depressive scale. However, conclusions about the relative performance of these techniques should not be made without the use of simulated data. This research study demonstrated the importance of expanding our techniques to explore what the data can tell us, as the findings would not have been revealed had the data only been explored by using hypothesis testing. Future research should include unsupervised statistical learning as a method to advance knowledge and classification in psychological research.

Description

Ph.D. University of Hawaii at Manoa 2016.
Includes bibliographical references.

Keywords

Citation

Extent

Format

Geographic Location

Time Period

Related To

Theses for the degree of Doctor of Philosophy (University of Hawaii at Manoa). Psychology

Related To (URI)

Table of Contents

Rights

Rights Holder

Local Contexts

Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.