BackHome
Refresh: OffView Personal Schedule
Session Summary

Sensitivity Analysis in Structural Equation Modeling

Sun, April 11, 10:40am to 12:10pm EDT (10:40am to 12:10pm EDT), Division D, Division D - Section 2 Paper and Symposium Sessions

Session Type: Demonstration/Performance

Abstract

Structural equation modeling (SEM) is used frequently in educational research for theory development and validation. Despite the sophistication of current methods available within the SEM framework, methods for sensitivity analysis to omitted confounders are scarce. Omitted confounders may lead to incorrect conclusions about theoretical relations involving latent variables as well as observed variables. This session demonstrates five new methods to quantify how strongly related an omitted confounder must be to variables in a structural equation model for the conclusions from the model to change. Therefore, these methods allow researchers to understand the robustness or vulnerability of the model to external misspecification. The steps of SEM sensitivity analysis methods are presented with applied examples from the existing educational research literature.

Sub Unit

Chair

Papers