This side session at the August 2023 AmeriGEO meeting in Costa Rica will teach attendees how to develop environmental-epidemiological models to predict the number of cases of a vector-borne disease such as dengue. The course will be taught in Spanish with simultaneous English translation. The R statistical language will be used to explore Generalized Linear Models (GLMs) and Random Forest (RF) models of the influence of environmental parameters on dengue incidence. Models like these can be important components of early warning systems, affording valuable time to preventative medical countermeasures and communication to reduce the incidence of disease. Datasets will be provided, but attendees who want to explore their own geo-referenced Dengue or other vector-borne disease data, are encouraged to bring it. Remote learning students (and any in-person students who plan to use their own computers) will be expected to arrive with their programming language and environment pre-installed and configured, including any required packages (see requirements below). The in-person course will be taught in a computer lab at the University of Costa Rica, so students without computers will have access to a pre-configured lab computer. Students unfamiliar with R are encouraged to review a quick primer in advance, and to attend the mini clinic sessions (see agenda below).
The skills and models developed in this class will be helpful starting points for developing more complex models for future purposes. An open dengue forecasting challenge (see example west nile virus forecasting challenge and prior dengue forecasting challenge) may be run several months after the AmeriGEO course, affording participants of the course an opportunity to continue to develop their models to be used in the challenge. All course materials will be open source and provided for future use, including for reference in the dengue forecasting challenge.
Instituto Gulich -Comisión Nacional de Actividades Espaciales (CONAE) Ximena Porcasi [[email protected]] Pablo Paccioretti [[email protected]]
09:00 - 11:00 Introduction to Dengue Early Warning and Climate Drivers
This 2-hour session serves as an introduction to the issue of managing dengue cases in Costa Rica. It will include presentations from the following individuals:
- Hunter Jones, U.S National Oceanic and Atmospheric Administration
- Jonathan Golden, U.S. Department of State
- Dr. Adriana Alfaro, El Ministerio de Salud de Costa Rica
- Karina Hernandez, Instituto Meteorológico Nacional de Costa Rica [pending]
- Dr. Ximena Porcasi, Comisión Nacional de Actividades Espaciales (CONAE)
11:00 - 13:00 AmeriGEO Closing Ceremony and Lunch
13:00 - 14:30 Mini Clinic 1 - R Basics
This 90-minute mini clinic will focus on basic R skills, including loading, wrangling, and plotting data, generating descriptive statistics, and the basics of r formula syntax.
15:00 - 16:30 Mini Clinic 2 - R Geospatial
This 90-minute mini clinic will focus on skills for working with geospatial data, such as environmental rasters and station-based data. It will also explore how to address common issues, such as spatial interpolation, in R.
16:30 - 17:30 Open Lab Time
Attendees are encouraged to explore the methods taught in the class using their own datasets. Instructors will be available to answer questions.
09:00 - 12:00 Generalized Linear Models in R
This session will cover a quick refresher on statistics through linear regression (distributions, assumptions, test statistics, diagnostics) through the use of generalized linear modeling packages in r. Advanced topics such as mixed effects models, cross-validation and information criteria will also be covered.
13:00 - 15:00 Random Forest in R
This session will cover the Machine Learning technique of Random Forests. Decision trees, bagging, boosting, and other techniques will be covered.
15:30 - 17:00 Open Lab Time
Attendees are encouraged to explore the methods taught in the class using their own datasets. Instructors will be available to answer questions.
All are welcome, but this course is primarily targeted at 2 audiences:
-
Working professionals who may be interested in developing operational dengue (and other vector-borne disease) forecasting capabilities, such as those at met services, space agencies, health ministries, and related entities.
-
Students in the Americas at the undergraduate and graduate level with some basic experience in programming and rudimentary knowledge of statistics.
-
Cross-train members of CR (and other national) ministries with needed skills for climate-linked VBD prediction to inform a pilot early warning system.
-
Develop skills of students and early career professionals in VBD forecasting in preparation for a VBD forecasting challenge to follow this training session.
-
Develop a variety of experimental models via the training and enhance them in the forecasting challenge to support an experimental, operational CR dengue forecasting system pilot.
-
Identify gaps in data, infrastructure, and other elements or technical needs that can be addressed over time.