Evaluating Psychometric and Imputation-Based Methods for Data Integration

Doctoral Dissertation

Abstract

IDA refers to combining data from independent studies into a concatenated dataset and analyzing the new data jointly (Curran & Hussong, 2009). IDA is an alternative to meta-analysis, which synthesizes summary statistics (such as effect sizes or standardized parameter estimates) from multiple studies. Pooling data for research is not a new idea; however, it has seen limited use in psychology. The IDA framework is innovative in its use of different models and methods to account for heterogeneity across different data sources. This dissertation evaluates the performance of particular applications of IDA in the field of behavior genetics, where meta-analysis is common practice. Behavior genetics researchers often perform exhaustive searches across the entire genome seeking genetic markers that are associated with an observed outcome. Because there are millions of genetic markers, and because even true gene associations have extremely tiny effect sizes, genome-wide searches are often combined across multiple studies to increase statistical power. Regression coefficients from tests of association are meta-analyzed, even though psychological or behavioral outcomes may be defined differently across studies. This dissertation tests the hypothesis that more precise measurement of an integrated outcome, achieved through IDA, can provide added power over typical meta-analysis in multi-study genome-wide searches. In chapter 1, a general overview of the IDA framework is discussed, genome-wide association studies are presented, and IDA for genome-wide searches is presented. Chapter 2 covers a simulation study evaluating measurement model IDA using a proposed bi-factor integration model. Chapter 3 covers simulation studies investigating multiple imputation IDA performance with a novel imputation model, namely, boosted decision trees as an extension of single-tree imputation. Chapter 4 presents an application of IDA to the Aggression in Children: Unraveling gene-environment interplay to inform Treatment and InterventiON strategies (ACTION) Consortium, a large-scale collaboration of six research projects studying the genetic underpinnings of aggression in children.

Attributes

Attribute NameValues
Author Justin M. Luningham
Contributor Gitta Lubke, Research Director
Degree Level Doctoral Dissertation
Degree Discipline Psychology
Degree Name PhD
Defense Date
  • 2018-03-30

Submission Date 2018-04-07
Subject
  • Integrative Data Analysis; Multiple Imputation; Genetic Consortia

Record Visibility and Access Public
Content License
Departments and Units

Files

Please Note: You may encounter a delay before a download begins. Large or infrequently accessed files can take several minutes to retrieve from our archival storage system.