Quadratic Nonlinear Path Analysis Modeling on Simulation Data

Abstract

This study aims to analyze quadratic nonlinear paths using the WLS method, perform hypothesis testing, and determine the best model between various sample sizes in the simulated data. The data used in this study is simulated data, where the data is applied to three exogenous variables, one endogenous variable, and one mediating variable with a correlation of 0.9 at various sample sizes (𝑛=100, 300, and 500). The results of this study indicate that there is variationin the various sample sizes that can affect the smoothness of the curve. The larger the sample size value, the better the resulting curve. The results of testing the linear parameter hypothesis show that there is a significant relationship between 𝑋1and𝑌1, 𝑋2and 𝑌1, 𝑋3and 𝑌1, and 𝑋1and 𝑌2at a correlation of 0.9 with a sample size of 100, 300, and 500. The best model is obtained in quadratic nonlinear path analysis with the largest 𝑅2value (95%) compared to the coefficient of determination of other models. The originality of this study is by using a quadratic nonlinear path analysis model and the useof simulation data which is rarely studied by other researchers.