Joint Species Distribution Models: Walking the walk or just talking the talk?
Details
Fitting joint species distribution models (JSDMs) that take into account associations between species often requires the use of more black-box style computational approaches that make evaluation beyond prediction challenging. This talk will stress test these models under various model mis-specifications and explain an approach to evaluating performance that focuses on inference. We assess performance according to metrics that are functions of the underlying species association matrix and that are often of interest in ecological contexts.
Agenda:
7:00 - Join early to network
7:10 - Opening announcements
7:15 - Dr. Sara Stoudt
8:00 - Open Q&A with the audience
8:15 - Breakout rooms for further discussion and networking
Speaker bio:
Sara Stoudt is a lecturer at Smith College and a statistician with research interests in ecology and the communication of statistics. Stoudt received her bachelor’s degree in mathematics and statistics from Smith College and a doctorate in statistics from the University of California, Berkeley. Her graduate work involved evaluating species distribution and abundance models under model mis-specification. While at Berkeley she was also a Berkeley Institute for Data Science Fellow and worked with Professor Deborah Nolan to teach statistical writing. Keep an eye out for their forthcoming book, Communicating with Data: The Art of Writing for Data Science. Follow her on Twitter: @sastoudt