The performance of parameter estimation in structural equation modeling under different statistical conditions

Abstract

The aims of this study were 1) to develop a model and simulate data under different statistical conditions, 2) to confirm model fit between developed models with the empirical data, and 3) to compare the factors influencing the efficiency of parameter estimation. The method, of this study was experimental research. The steps were as follows: 1) data generated by using the Monte Carlo method for analyzing the structural equation model 2) analyzing the structural equation model and confirming model fit withindices for all 36 models 3) comparing the factors influencing the efficiency of parameter estimation. The factors consisted of sample size, parameter estimation methods, and Level of Multicollinearity. The results showed that 1) the simulation conformed to the conditions 2) model fit indices showed that using Maximum likelihood (ML) and weighted least square (WLS) estimation methods, the model was consistent with the empirical data at all levels of sample size and found that the chi-square statistic tended to increase when the sample size was larger; and 3) In a comparison of factors affecting the efficiency of parameter estimation from the relative bias index (RB) and Monte Carlo standard error (MCSE), there was an interaction between the sample size and the parameter estimation method that affecting the efficiency to parameter estimation at a significance level of 0.05. Multiple comparison to explore the mean differences between pairs of groups shows that at the 200 sample size, the ML vs. Generalized least square (GLS) and ML vs. WLS estimation methods showed a statistically significant difference in both RB and MCSE indices. Therefore for sample sizes 400 and 1000 sample, parameter estimates were no different.