Four-Way Interaction Effects on the Major Depressive Disorder Based on Multifactor Dimensionality Reduction Method

• Activity Number: 188 - Contributed Poster Presentations: Section on Nonparametric Statistics
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Jung Yeon Lee* and Wonkuk Kim and Judith S Brook
• Companies: NYU School of Medicine and Chung-Ang University and NYU School of Medicine
• Keywords: Multifactor dimensionality reduction; 4-way interaction; Major depressive disorder; Longitudinal data

Abstract

The adverse consequences of Major Depressive Disorder (MDD) include unemployment, life functioning disability, and impairments on executive function. To prevent MDD, a number of research papers examined the precursors of MDD using logistic regression analyses. Single and two-way interaction effects of the precursors on MDD have been reported. The current study, however, conducts the multifactor dimensionality reduction (MDR) method rather than logistic regression. 674 participants were recruited from the 5-wave Harlem Longitudinal Development Study (53% African Americans and 47% Puerto Ricans; 60% females). In order to detect higher order nonlinear interaction effects among the predictors of MDD, we applied the MDR method. Using the MDR analysis, four-way interaction effects of poor health conditions at age 24, cigarette use at age 29, cannabis use at age 29, and body mass index at age 29 were identified with 69.36% testing prediction accuracy and 10 out of 10 cross-validation consistency. In contrast, the logistic regression analysis often fails to detect higher-order interactions. The odds ratios of some interesting configurations among the four predictors were also calculated.