Modeling nonignorable missing responses in cognitive diagnostic models

Xiaomin LI, Wen Chung WANG, Kuan Yu JIN

Research output: Contribution to conferencePaper

Abstract

In educational and psychological assessment, missing data often occur and usually are not ignorable. The non-ignorable missing data mechanism must be modeled to reduce the biased parameter estimates. This study aims to develop a class of cognitive diagnostic models (CDMs) to account for non-ignorable missing data. By forming a joint model for the response data and missing data, the missing mechanism can be well considered under CDMs. Simulation results showed that, parameters were recovered fairly well with missing models. Although treating missing data as ignorable did no harm in the item parameter estimation, it yielded lower correct classification rates on the latent profiles than when the missing data mechanism was properly considered.
Original languageEnglish
Publication statusPublished - Apr 2014

Fingerprint

Parameter estimation

Citation

Li, X., Wang, W.-C., & Jin, K.-Y. (2014, April). Modeling nonignorable missing responses in cognitive diagnostic models. Paper presented at the 2014 AERA Annual Meeting, Philadelphia, Pennsylvania.