Psychometric Analysis of a Helping Scale with the Graded Response Model

Abstract

This article presents the results of the application of the graded response model (GRM) to the items of the Helping Scale, an eight-item test which measures helping behavior, a type of prosocial behavior. A convenience sampling strategy was employed, which made it possible to enroll 1383 Argentine university students (80% women). 73.5% of the participants attended the University of Buenos Aires, 18.9% attended other national public universities, and 7.6% attended private universities in the City of Buenos Aires. One of the items, which referred to altruism, presented gender non-uniform differential functioning and was removed from the scale. The GRM assumptions of local independence and one-dimensionality were successfully corroborated with the chi square LD index and confirmatory factor analysis, respectively. After that, the GRM was implemented. Average threshold parameters tended to be located at middle and high levels of the trait scale, while slope parameters displayed high item capacity to distinguish between participants with different levels of helping behavior. The analysis of the information function of the test showed acceptable accuracy in a wide range of helping levels, encompassing most people. Reliability was also adequate. One item, which refers to the academic domain, exhibited a low and flat level of information and out-of-range threshold parameters; therefore, it will need to be reformulated.