Regression Model Selection Analysis of Methanol Conversion Based on Temperature, Residence Time, Concentration, Oxygen Ratio, and Reactor System

Abstract

This study aims to determine the best regression model that explains the effect of temperature, residence time, methanol concentration, oxygen to methanol ratio, and reactor system on methanol conversion in supercritical water. Preliminary analysis showed a violation of the multicollinearity assumption, which affected the validity of the multiple linear regression model. To overcome this and determine the optimal model, variable selection was performed using the stepwise selection method. This method was evaluated based on predictive power, model accuracy and statistical validity. The results showed that the stepwise method produced an optimal model in predicting conversion. Reactor system and temperature were the most significant variables affecting methanol conversion. The conclusion of this study shows that the variable selection approach with stepwise selection can be effectively used to identify the best regression model, when classical assumptions are met. These findings make an important contribution to the optimization of supercritical water-based chemical processes.