The Impact of Conditional Dependency and Its Detection on Cognitive Diagnosis Modeling
Kevin Carl Santos · Proceedings of the 2019 AERA Annual Meeting · 2019
Estimating cognitive diagnosis models (CDMs) typically assumes conditional independence.This study investigates how violating this assumption affects the validity of inferences obtained from CDMs, and how it can detected.Results indicate that conditional dependency yields biased estimates and understated standard errors, resulting in higher classification accuracy and item discrimination indices.