Date of Award
Master of Public Health (MPH)
Background- Many different mathematical models are used to assess and predict the outbreaks. The model is selected by the characteristics of the outbreaks. Here, we utilize the generalized growth model (GGM), one of the simplest mathematical models, with the real outbreaks to compare two parameter estimation methods.
Materials and Methods- 25 outbreaks are used to analyze. We use GGM with the ascending phase of each outbreak and estimated the r and p parameters with both the least square (LSQ) and maximum likelihood estimation (MLE) methods. For both parameter estimation methods, we conduct the parametric bootstrap method to construct the confidence interval of parameters. We compare the two estimation methods by the RMSE, Anscombe residual, and prediction coverage.
Results- The result shows that most outbreaks have similar r and p parameters, RMSE, Anscombe, and prediction coverage for LSQ and MLE. Although Anscombe values for LSQ are higher than the values for MLE, the difference between results of the two methods are minimal for the most outbreaks.
Conclusion- The study is shown that LSQ and MLE do not result in different values of the parameter estimation, RMSE, Anscombe, and prediction coverage with GGM.
Lee, Yiseul; Roosa, Kimberlyn M.; Tariq, Amna; and Chowell, Gerardo, "Systematic Comparison of Parameter Estimation Approaches Using the Generalized-growth Model for Prediction of Epidemic Outbreaks." Thesis, Georgia State University, 2019.
Available for download on Wednesday, May 06, 2020