Visualization Tools for Backward Elimination Technique in Multiple Regression Time Series Modelling of CO2 Emissions in Malaysia

Understanding multiple regression time series modelling is crucial because the procedures involve intricate statistical methods. This study incorporates a flowchart that clearly illustrates the steps for modelling a response variable affected by several explanatory variables via the backward elimination technique. The first objective of this study is to utilise ten graphical tools, comprising charts and tables, for visual assessment to support formal evaluations in model diagnostics using R programming. The aim is to provide comprehensive insights and improve the overall understanding of the modelling procedures. The visualisation tools include criteria for multicollinearity, goodness-of-fit, and underlying assumptions of normality, homoscedasticity, zero serial correlation, and volatility in the residuals. The second objective involves implementing modelling procedures to obtain a well-specified model in a real-world context, demonstrating its practical value and implications. In this instance, the selected response variable is carbon dioxide (CO2) emissions, significantly contributing to global warming. In Malaysia, CO2 emissions increased continuously from 1990 to 2022, with an alarming average annual growth rate of 4.9%. The visual diagnostics have helped guide the elimination of some explanatory variables in the initial model and refined the models, resulting in a well-specified final model that is parsimonious and explains 98.6% of the variability in CO2 emissions. The final model suggests that high fossil fuel use and GDP per capita are contributing factors to increased CO2 emissions in Malaysia. The study recommends government action and investment in renewable energy to reduce CO2 emissions by 45% by 2030 and achieve net-zero emissions by 2050.