The Canadian Chapter is hosting a week-long short course July 24-28, 2023 titled “FORECASTING FOR DECISION-MAKING: AN EPIDEMIOLOGICAL & ECOLOGICAL PERSPECTIVE”
This 5-day Summer School will provide participants with lectures and hands-on experience in forecasting epidemiological and ecological phenomena in decision-making contexts. The main objectives of this short course are to improve participant capacity to (1) produce more reliable and reproducible forecasting models; (2) generate accurate forecast horizons and relevant scenario analyses; (3) integrate decision-makers’ priorities and interests into forecasting models; and (4) effectively communicate foresting model results, assumptions and uncertainties across various audiences including decision-makers and the public.
Each day will consist of an interactive lecture component in the morning and small group projects in the afternoon. The lecture component will combine technical lectures to address specific challenges (e.g., tools to facilitate workflow; how to characterize underlying distributions) and case studies that illustrate mismatches between common perceptions and reality in using forecasting to inform policy (e.g., the best forecasting model may not be the one used for decision-making). In the afternoon sections, participants will use open-source datasets provided by government agencies (e.g., Ontario Ministry of Health; Environment and Climate Change Canada) to engage with forecasting questions that are of priority and interest to decision-makers. By using government datasets and having participants generate forecasts that correspond to proposals outlined by government entities, we will provide a unique context for participants to troubleshoot, iterate, and refine their models to successfully meet forecasting and decision-makers requirements. In practice, a diversity of data sources may be used for forecasting; and as modelers, we often use data collected, cleaned, and organized by other teams. These data reflect people and communities, and in this workshop, we plan to provide an introductory discussion, via a guest lecture, on meaningful engagement with data under a framework of data justice and data sovereignty. Finally, as forecasting is increasingly being adopted as a viable tool to inform decision-making across various disciplines, in addition to epidemiological forecasting examples, we also plan to have examples and datasets from Environment Climate Change Canada to provide a cross-disciplinary perspective for how to tackle common forecasting and decision-making challenges.
The short course will be held in coordination with both the Mathematics for Public Health at the Fields Institute for Research in Mathematical Sciences and the Ecological Forecasting Institute in the United States, who will provide access to a pre-course online forecasting training. The pre-training course will ensure all participants share a basic and common knowledge of forecasting. Additionally, the pre-course training requirements will also allow for participants of all skill-levels to attend, thus truly contributing to capacity building. We plan to have two government datasets available: Covid-19 datasets to use as an epidemiological case-study for groups to forecast, as well as long term wildlife monitoring data for an ecological case-study. In addition to the in-person format, all sessions will be streamed and recorded through Zoom. Teaching assistants or instructors will be assigned to monitor Zoom and address questions posed in the online chat. This will ensure a broader reach as well as allow for the content of lectures, hand-on exercises and datasets to be made publicly available beyond the duration of the short course.
Pre-course requirements: Pre-recorded and introductory videos to forecasting, in collaboration with the Ecological Forecasting Initiative.
Optional tutorials: We plan to host 5 weekly sessions prior to the course start on topics useful for generating forecasts. Topics will include: Bayesian Analyses in R/JAGS, Getting started with Github, Building and Fitting SIR Models, Uncertainty Propagation, and State-Space Models.