Factor Structure Equivalence across Gender of a Volition Model among Engineering Undergraduates: Comparison of LRT, ΔCFI, Gamma Hat and McDonald’s NCI Approaches

Abstract

Test of equal performance of a structural or measurement model in measuring a psychological phenomenon or a construct, in subjects across multiple groups, technically known as measurement invariance or measurement equivalence testing, is an essential practice in Psychometrics, which is not commonly followed in the Indian context. To promote awareness and encourage initiation of its inclusion in tool validation exercises, measurement invariance testing of the volitional model of self-regulated learning varying with respect to gender was conducted using 533 (373 males and 160 females) second and third year computer science and mechanical engineering students of the Punjab state of India. The instruments of data collection were the Academic Procrastination Scale Short Form by Yockey (2016), the Zimbardo Time Perspective Inventory Short Form by Orosz et al., (2017) and the Academic Delay of Gratification Scale by Bembenutty and Karabenick (1998). Apart from the popular approach of measurement invariance testing like Likelihood Ratio Test (LRT), lesser known invariance testing techniques like Cheung and Rensvold (2002), Steiger’s Gamma’s Hat and McDonald’s NCI criterion which are unaffected by model complexity, sample size and overall fit measures, were administered on the collected data. The model showed acceptable overall CFA fit. Configural invariance was supported across gender. Metric invariance was rejected by LRT, ΔCFI, and Gamma Hat but supported only by McDonald’s NCI by a small margin. The same result was confirmed by sensitivity analysis too. Therefore, the model should be interpreted as configurally invariant but not robustly metrically invariant across gender. Implications of the study with respect to engineering education and psychometric practices are discussed.