The Implementation of Robust Parameter Estimation in Soil-Borne Infectious Disease Transmission Model through Winsorization Approach

In epidemiology, applied scientists often construct models of complex dynamic processes using ordinary differential equations (ODE) which provide a simple representation of the spread of an infectious disease. One of the essential aspects of modelling is to define parameters in an outbreak model. It is interesting but challenging to estimate ODE parameters especially when the data contains outliers. The robust estimator based on Winsorization is recommended when dealing with the presence of outliers. This approach had been proposed as a solution for treatment of data when the outlier exists, and distribution is non-normal in identifying parameters. In this paper, we propose a robust parameter estimation method to identify the incidence rate in soil- borne plant disease transmission model by utilizing the Winsorization method. Robust estimators, namely the Winsorized mean, replace the arithmetic mean used in this study. Different Winsorization percentages of 5%, 10%, 15% and 20% are regarded for comparative purposes. A simulation analysis was conducted to examine the efficiency of the model using the proposed method in terms of mean square error (MSE). The results revealed that the 5% Winsorized mean method provides a better estimate for an incidence rate in the presence of outliers which gives the smallest MSE. The newly constructed model with robust parameter estimation substantially enhanced the performance of the soil-borne plant disease transmission model in identifying incidence rate and resolving initial forecast errors, as well as providing greater accuracy for long- term predictions. Future research may consider expanding the robustness of parameter estimation using novel and advanced statistical methods, as well as improving the accuracy of statistical analysis through the advancement of precise prediction of soil- borne plant disease transmission model.