The Effect of Items Direction and The Number of Response Categories on Differential Item Functioning Using Graded Response Model

Abstract

This study examined the effects of item wording direction (positive vs. negative) and response category formats on differential item functioning (DIF) across gender, using the Graded Response Model (GRM). Six versions of the Rosenberg Self-Esteem Scale, systematically varied by item wording and response categories (3, 4, and 5-point), were administered to a random sample of 7,200 teachers. Results indicated that negatively worded items were consistently more susceptible to gender-based DIF, particularly as the number of response categories increased. Specifically, DIF was observed in two items on the three-point scale, three items on the four-point scale, and all five negative items on the five-point scale. In contrast, positively worded items showed minimal evidence of DIF across response formats, with only one instance of non-uniform DIF observed on the five-point scale. These findings emphasize the role of item wording and response format in contributing to DIF and support the use of three response categories when combining positive and negative items to reduce DIF occurrence