STATISTICS SEMINAR – Feb. 11
Join Dr. Andrew MacDonald, a statistical ecologist based in Montreal for their talk on Thursday, February 11th about “Wild-Caught Data and Half-Tame Models: Applied Statistics Out in the Field”.
ABSTRACT:
The jungle is a tough environment for statistics. Data is messy, incomplete, and full of measurement error. Models based on domain expertise often clash with textbook statistical models. In this talk I’d like to share some experiences from my work as a community ecologist. This will cover a few loosely connected examples. First, our recent work building a bayesian model to predict the connections in ecological networks. This model makes better predictions than previous attempts and matches basic food web theory more closely. Second, a discussion of validating a model before fitting via prior predictive checks. This approach can be a better guide to setting a “prior” for many biologists, as well as a good aid to defining and sampling an ecological model. Finally I want to share some insights from teaching applied statistics to grad students in a course-based program. In this course I wanted to experiment with teaching a recent approach to statistical workflows: the “tidymodels” packages in R. I think this raises a lot of interesting questions about how statistics is taught, understood, and practiced.
DATE: Thursday, February 11th, 2021
WHERE: Zoom (see below for more information)
WHEN: 2:30pm – waiting room will open at 2:15pm.
If you wish to attend a Statistics Seminar please contact Po Yang (Po.Yang@umanitoba.ca) from the U of M’s Statistics Department for more details and Zoom meeting information.
ADDITIONAL INFORMATION:
Andrew MacDonald is the Training coordinator of a NSERC CREATE program called BIOS2 . He teaches applied statistics at the graduate level and participates in several synthesis groups, where his main interest is in mechanistic models of community composition of ecological communities. Right now he is most interested in how to use Bayesian methods to measure and represent uncertainty in ecology, as well as reproducible workflows that allow scientists to communicate clearly with each other and with their audience.
Andrew will be available for one-on-one or lab group meetings on February 11th. As is common in many biology departments, a (virtual) lunch with graduate students will be scheduled on February 11th to have an informal discussion with Andrew about research, life as an applied statistician, or anything in between. We hope many of the graduate students will take this opportunity, it’s a nice chance to hear the perspective from someone else and also a chance to socialize with other students. Please email Aleeza.Gerstein@umanitoba.ca if you are interested in either a meeting for you or your group or if you’d like to attend lunch.
