Multiple Inference with Applications in Social Science

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Operations, Information and Decisions
Discipline
Statistics and Probability
Subject
applied statistics
Bayesian statistics
multiple hypothesis testing
social science
Funder
Grant number
License
Copyright date
2023
Distributor
Related resources
Author
Bowen, Dillon
Contributor
Abstract

Multiple inference - comparing many "things" simultaneously - is common in social science. Behavioral scientists regularly compare the effects of many behavioral interventions, personality psychologists study individual differences by comparing many people, and diversity researchers compare the impact of policies on many demographic groups, to name only a few examples. Unfortunately, it is difficult to perform multiple inference correctly, and many social scientists do not. As a result, social scientists often draw unfounded or misleading conclusions from their data. This thesis addresses the problem of multiple inference, especially in social science. Chapter 1 describes several of the best-performing multiple inference techniques - including several new techniques developed in later chapters - and shows how applying them to social science datasets can lead to substantially different conclusions compared to traditional statistics. Chapters 2 and 3 describe the new multiple inference techniques introduced in Chapter 1 (inference after ranking, Bayesian ranking, and Bayesian selection) in greater detail. Importantly, we provide an open-source statistics package so that social scientists and other researchers can apply correct multiple inference techniques in a few lines of code.

Advisor
Simmons, Joseph
Tetlock, Philip
Date of degree
2023
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation