manuscript

Abstract

Forced choice (FC) is one of the most used forms measurement for non-cognitive assessments, which can effectively resist faking and some other response biases compared to the Likert-types scales, and has been a popular topic in the field of industrial organizational psychology in recent years. Inspired by Lee et al., (2019) study, the present study proposed a 2PL-RANK model as a variant of the GGUM-RANK for fitting dominance RANK items. To improve the efficiency of parameter estimation, the authors apply the stEM algorithm to the 2PL-RANK model, which greatly improves the efficiency of parameter estimation in joint estimation. What’s more, we derived information functions for this model based on the logic of Joo et al., (2018). Then, simulation studies were conducted to examined the recovery of model's parameters with RANK triplet responses, which manipulated four factors, with sample size, the number of dimensions, the number of blocks measured in each dimension, and the correlation between dimensions. Results show that the 2PL-RANK model performed well in estimating item and trait parameters. Finally, the utility of 2PL-RANK and Thurstonian IRT model (TIRT) in a 24-dimensional FC personality test was compared. An empirical study was then conducted based on a 24-dimensional FC personality test to illustrate the practical use of the model.