Identify a Subset of Items That Predict the Total Score of a Psychological Assessment Measure

• Activity Number: 374
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Tzu-Cheg Kao* and James A. Naifeh and Carol Fullerton and Robert Ursano
• Companies: Uniformed Services University of the Health Sciences and Uniformed Services University of the Health Sciences and Uniformed Services University of the Health Sciences and Uniformed Services University of the Health Sciences
• Keywords: mutilation fear ; posttraumatic stress ; anxiety ; random forest

Abstract

Although psychological assessment measures are valuable clinical research tools, the length of many instruments increases administration time, which limits the scope of survey research and places additional burden on research participants. It would be beneficial to shorten existing measures without substantial loss of information. Using data from 770 active duty U.S. Army soldiers (694 male, 68 female, and 8 did not report gender), we intend to use random forest, a non-parametric machine learning method, to identify a reasonable subset of items that predict the total score of the 30-item Mutilation Questionnaire (MQ). The MQ is a self-report measure mutilation fear, which a cognitive-emotional vulnerability found to be associated with risk for anxiety and posttraumatic stress disorder. Further statistical results will be presented later.