Direction dependence analysis in latent variable contexts : comparison of normal and non-normal item response theory models
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Thesis
Subject
Abstract
One common goal of science is to uncover causal relationships. When randomized controlled trials (RCTs) are not feasible, researchers often rely on survey-based observational data to explore causal relationships. In observational research, most used statistical methods are based on covariances and correlations and assume that the collected data are normally distributed. However, correlation does not imply causation, and research has repeatedly shown that data in social sciences often deviate from normality. This study focuses on a method called Direction Dependence Analysis (DDA), which is a confirmatory approach designed to determine the causal relationship between two variables that make use of non-normality. This study aims to explore the effect of psychometric properties of survey data on DDA through two Monte-Carlo simulation studies, each involving different psychometric models and scoring approaches with and without the presence of hidden confounders. A real-world example highlights how DDA can be applied to survey-based data using different psychometric models and scoring methods. The results showed that the number of response categories, instrument length, and sample size have positive effects on correctly identifying the true causal model. Additionally, Davian-Curve Item Response Theory scores outperformed other scoring methods. Recommendations for applied research are given, and limitations and future directions are discussed.
Table of Contents
PubMed ID
Degree
Ph. D