Collider Bias: Exploring Fidelity of Implementation in Causal Inference Frameworks

• In Event: Innovations in Experimental Design and Causal Inference

Abstract

Treatment fidelity is an important feature in ensuring internal and external validity
in intervention studies. Despite the significance of treatment fidelity, there is a lack of
research examining the role of fidelity as a collider in the analysis of treatment effects.
This study is an investigation of fidelity of implementation as a collider in casual inference
frameworks. Using simulated data, this study detected how a collider bias undermines the
effect of a treatment on an outcome. The results revealed that adjusting for fidelity as a
collider induced a spurious association between two parent variables, tutor support and child
task orientation. This study will contribute to existing studies of collider bias, and suggest
optimal ways to mitigate the bias.

Authors

• Hyeri Hong, California State University - Fresno

• Laurentius Ardian Susadya, University of Iowa