Description & Objective
In this course, we will cover the steps and decisions involved in calibrating a mathematical model in R. We will begin the course with an overview of when model calibration is necessary and will introduce a general model calibration framework. We will then engage students in an extensive hands-on exercise where they will implement the calibration of a simple mathematical model in R using a simple random search. We will then introduce more advanced calibration approaches, including Latin hypercube sampling, directed search algorithms (e.g., Nelder-Mead), Bayesian calibration, and other iterative calibration approaches (e.g., genetic algorithms). We will discuss the tradeoffs of different calibration approaches and will identify scenarios when one approach may be more appropriate than others.
At the end of the course, participants will be able to:
- Understand the choices that must be made when calibrating a model
- Correctly interpret the results of a model calibration and visualize these results in R
- Implement the calibration of an existing model in R using any of the following approaches:
- Random search
- Latin hypercube sampling
- Directed search with the Nelder-Mead algorithm
- Bayesian calibration using the Metropolis-Hastings algorithm
- Genetic algorithms
- Understand the strengths and weaknesses of different calibration approaches
All R programming templates used in this short course will be provided to participants after the course for future use.
Course Director
Eva A. Enns, MS, PhD Division of Health Policy and Management, University of Minnesota School of Public Health, Minneapolis, MN
Course Faculty
Fernando Alarid-Escudero, PhD Health Policy and Management, Division of Public Administration, Center for Research and Teaching in Economics (CIDE), Aguascalientes, AG, Mexico