Methods to Retrofit and Validate Q-Matrices for Cognitive Diagnostic Modeling

Cognitive diagnostic models (CDMs) are a family of constrained latent class models that estimate relationships between observed item responses and latent attributes (Rupp and Templin, Educ Psychol Meas 68:78-96, 2008). An important input needed in any CDM is the Q-matrix, an item-by-attribute table that represents a particular hypothesis about which attributes are required to answer each test item successfully. A large number of CDMs have been developed; however, many applications involve retrofitting a CDM to an existing non-diagnostic test. In this study, we conducted a systematic review to describe the current picture of retrofitting Q-matrices to non-diagnostic tests and consequently using the tests for diagnostic purposes.