APS
2024 APS Annual Convention · 2024
Inference on the Indirect Effect in the Presence of Missing Data: The Bias-Corrected Bootstrap Confidence Interval Results in Inflated Type I Error Rates When Implemented with Multiple Imputation Procedures
- Tristan Tibbe
University of California, Los Angeles - Amanda Montoya
University of California, Los Angeles - Catherine Crespi
University of California, Los Angeles - Craig Enders
University of California, Los Angeles
Abstract
Two different multiple imputation methods for mediation analysis on incomplete data implement the bias-corrected bootstrap confidence interval to determine the significance of the indirect effect. Our simulation study shows that this approach can result in elevated type I error rates, and alternative inferential techniques offer greater type I error control.
Other
