EVALUATING THE EFFECTIVENESS OF THE SIMULTANEOUS ITEM BIAS TEST IN DETECTING DIFFERENTIAL ITEM FUNCTIONING ACCORDING TO GUESSING PARAMETER VARIATION AND SAMPLE SIZES

Abstract

The Simultaneous Item Bias Test (SIBTEST) compares the performance of different groups on individual test items, considering their overall ability levels. It provides valuable insights into the potential sources of bias in the assessment and aids in making informed decisions regarding item inclusion or modification to enhance the fairness and validity of the test. The current study aims to assess guessing parameters and sample size on the effectiveness or performance of Statistics of the Simultaneous Item Bias (SIBTEST) as a method to detect differential item functioning (DIF) according to a 3-Parameter Logistic Model (3PLM) using Item Response Theory (IRT). The study adopted the experimental methodology, which controls the variables. The study participants are individuals with ability levels (θ) randomly generated from a standard normal distribution N (0,1). WinGen-3 program was used to generate data, and SIBTEST software was used for calculating the statistic (β) value. To achieve the study objectives, data have been generated for a (50) item test for both the reference and the focal groups. Three levels of guessing parameters (0.10, 0.15, and 0.20) were chosen for the focal group, whereas the guessing parameter of the reference group was stabilized at (0.20). Three sample sizes were used (250, 500, and 1000) individuals. To judge the effectiveness of statistics (β), both type one error and Test power were employed as measurements. The findings revealed that the test power ratio increases when the difference in the value of the guessing parameter decreases between the two groups with the increase in sample size. Moreover, the type one error was within the defined significance level (α < 0.05) when the value of the guessing parameter was approximated between the two groups, with an increase in sample size.