ESTIMASI BAYESIAN PADA MODEL PERSAMAAN STRUKTURAL DENGAN VARIABEL KATEGORIK TERURUT

Abstract

This article explains about parameter estimation of structural equation model with ordered categorical variable using Bayes method. The basic assumptions of SEM are the data type is continuous, minimum scale is interval, and it has to satisfy the normality assumption. The categorical data is ordinal data which the observation is in discrete form, and to treat the categorical data as normally distributed continuous data is by finding threshold parameter for each categorical data. Bayes method only focuses on individual data by combining sample data and the research data before (prior information), in order to minimize the error rate. Hence, the parameter estimation of structural equation model can be obtained well. In this estimation process, it is done numerically by using Monte Carlo method, i.e. Gibbs Sampling and Metropolis Hasting.