Enacting Alternating Least Square Algorithm to Estimate Model Fit of Sem Generalized Structured Component Analysis

Abstract

Structural Equation Modeling (SEM) is a statistical modeling technique that combines three methods, namely factor analysis, path analysis and regression analysis to test a theoretical model in social science, psychology and management. Covariance-based SEM is a parametric SEM that must meet several parametric assumptions such as, multivariate normally distributed data, large sample sizes and independent observations, so that, variance-based SEM was developed to overcome the problem of covariance SEM, namely the Generalized Structured Component Analysis (GSCA) method. This study aims to implement the GSCA method on factors data that are expected to have an effect on the level of behavioral intention towards online food delivery services and to examine the significance of the mediating variable on the structural relationship. The results of hypothesis testing with a 95% confidence level showed that the quality of convenience motivation, prior online purchase experience, and attitude towards online food delivery services had a significant effect on behavioral intentions towards online food delivery services. The fit value is above 0, 523 which indicates that the model is able to explain around 52,3% of the variation of the data. Furthermore, the hedonic motivation variable has a significant effect on convenience motivation. Post usage usefulness and prior online purchase experience variables significantly affected the attitudes towards online food delivery services. The proposed model using GSCA achieves a much better result (good fit) compared with the previous model using Confirmatory Factor Analysis (CFA) with marginal fit.