Applications of Advanced Statistical Modeling Techniques for Understanding Depression
Abstract
We conducted three investigations into different facets of major depressive disorder. Chapter One describes the application of hierarchical modeling techniques in combination with the Linguistic Inquiry Word Count to characterize associations between depressed individuals’ emotion regulation strategies and their discrete state emotion. State emotion was associated with emotion regulation strategy selection, independent of the adaptiveness of the coping response. In Chapter Two we investigated the relationship between urban living and depression, using meta-analytic and meta-regression techniques to characterize how this relationship has changed over time across both developed and developing countries. In developed countries, depression prevalence was higher in urban areas than rural areas. In developing countries, no overall effect of urbanicity was found, although a relationship between urbanicity and depression appears to be emerging over time. Chapter Three describes an investigation comparing the accuracy of traditional and machine learning model building algorithms for predicting patient outcomes in response to antidepressant treatment. Traditional model building techniques designed for inferential analysis were found to be poorly suited for the purposes of prediction. Machine learning techniques designed for prediction can make accurate predictions; however, their utility is limited by how they are applied and to their access to informative predictors.
Subject Area
Psychology|Clinical psychology
Recommended Citation
Xu, Colin, "Applications of Advanced Statistical Modeling Techniques for Understanding Depression" (2022). Dissertations available from ProQuest. AAI29259586.
https://repository.upenn.edu/dissertations/AAI29259586