Rasch tree modeling for DIF analysis: how sample and item characteristics influence DIF detection

Identifying differential item functioning (DIF) is an important component in evaluating measurement procedures for evidence of validity and fairness. The Rasch Tree model provides researchers with a data-driven and exploratory approach to identifying DIF that incorporates examinees' simultaneous classifications with respect to multiple covariates (e.g. demographic groups). Although the Rasch Tree approach is promising, research is limited concerning the impact of sample and instrument characteristics. We used a real data analysis and simulation study to systematically examine the effects of sample size, subgroup differences, and the number of items containing DIF on the accuracy of DIF detection with the Rasch Tree DIF approach. Sample size, DIF presence, and covariate balance all significantly affected node selection within the Rasch Tree approach to DIF detection. We discuss implications for research and practice.