Evaluating Ten Simple Regression Models against a New Model: A Comparative Performance Study

Evaluating regression models is very important in statistics, data science, and research because it ensures that the model is not just fitted but also meaningful, reliable, and useful for decision-making. This study compared the performance of ten simple regression models with a newly developed model using Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and Hannan-Quinn Information Criterion (HQIC) as evaluation metrics. The analysis utilized two datasets: a real-world dataset from 60 patients at Federal Medical Centre, Owerri, Imo State, Nigeria and simulated datasets of varying sizes (10, 20, 30, 50, 100, 500, 1000, 3000, and 5000) to validate the results. The study evaluated ten simple regression models, including Arctangent, Sinusoidal, Square root, Exponential, Power, Hyperbolic, Logarithmic, Polynomial, Quadratic and Linear models, against the proposed model using the R-Studio package. The results showed that the developed model outperformed all other models, with significantly lower AIC, BIC, and HQIC values. The Wilcoxon Signed-Rank Test confirmed that these differences were statistically significant, indicating the developed model's superior performance. The study concluded that the developed model demonstrated robustness and superior performance across different sample sizes, making it a reliable choice for future applications. The findings suggested that the developed model could be a valuable tool for modeling and analysis in various fields.