Item Response Theory Assumptions: A Comprehensive Review of Studies with Document Analysis

Item Response Theory (IRT), over its nearly 100-year history, has become one of the most popular methodologies for modeling response patterns in measures in education, psychology and health. Due to its advantages, IRT is particularly popular in large-scale assessments. A pre-condition for the validity of the estimations obtained from IRT is that the data meet the model assumptions. The purpose of this study is to examine the testing of model assumptions in studies using IRT models. For this purpose, 107 studies in the National Thesis Center of the Council of Higher Education that use the IRT model on real data were examined. The studies were analyzed according to sample size, unidimensionality, local independence, overall model fit, item fit and non-speedness test criteria. According to the results, it was observed that the unidimensionality assumption was tested at a high level (89%) and Factor Analytic approaches were predominantly used. Local independence assumption was not tested in 36% of the studies, unidimensionality was cited as evidence in 40% of the studies and tested in 24% of the studies. Overall model fit was tested at a moderate level (51%) and Log-Likelihood and information criteria were used. Item fit and Non-Speedness testing were tested at a low level (26% and 9%). IRT assumptions should be considered as a whole and all assumptions should be tested from an evidence-based perspective.