Predicting monkeypox infection ratios using Bayesian and maximum likelihood methods of panel data: Employing the weighted exponential regression model

Abstract

This study employs advanced statistical techniques, namely Maximum Likelihood Estimation (MLE) and Bayesian inference, to estimate the parameters of a weighted exponential regression model for panel data, accounting for both fixed and random effects. The empirical analysis utilizes monkeypox incidence data from the Americas, Africa, and Europe over the period 2022–2023, complemented by simulated datasets of varying sample sizes (15, 30, 45, and 60) to thoroughly evaluate the model’s performance. A comparative analysis using the Mean Absolute Percentage Error (MAPE) criterion reveals that the Bayesian estimation method for random effects outperforms both the fixed effects model and the MLE approach for both fixed and random effects. Additionally, the model is used to forecast monthly infection rates over the next six months, with the Bayesian random effects model indicating a significant decline in infection rates, approaching near-zero values. This highlights the Bayesian framework’s ability to accurately capture and predict the dynamics of the weighted exponential regression model in panel data contexts involving random effects.

Journal of Statistical Research 2025, Vol. 59, No. 2, pp. 167-181