The Effectiveness of Bootstrapping Procedure in Estimating Person-Fit Statistic Cutoff- Score for Polytomous Items in Noncognitive Scales

Abstract

This study aimed to assess the effectiveness of bootstrapping procedure in estimating  person-fit statistic cutoff score (denoted as ) to detect aberrant responses in noncognitive scales. To achieve the aim of this study,  a sample of 1000 students was selected from the third and fourth year undergraduate students at Yarmouk University (N= 15152) using multi-stage sample procedures. A questionnaire consisted of 62 Likert Scale items designed to assess the quality of academic programs from students perspectives was used to collect the data. The datasets were analyzed utilizing graded response model of IRT. The results showed that person-fit statistic distributions are not normally distributed  and still have negative skewness and positive kurtosis. Moreover, the results showed that using cutoff scores calculated by bootstrapping procedures () to categorize responses to fit or misfit is preferred asdetected more aberrant responses when it is compared with Z critical value at the same level of statistical significance