This is for those who are planning to join the StanCon 2024 model selection tutorial.
As I have already several recorded talks about details of some model selection methods like cross-validation, in this tutorial the plan is following:
- Higher level overview on model checking, model selection, and hypothesis testing with pointers to videos and other material with more details.
- Hands-on session, in which the participants can play with some of the existing case studies or work with their own data and models, with I, @n-kall and Teemu Säilynoja going around answering questions.
- We pick the most interesting examples from the participants, and present to all our advice how to proceed with model selection in those case. The rest of the time is used for Q&A.
So bring your laptop (fully charged), models and data to the tutorial.
You may also reply here for topics you would like to here more in the overview part or in the Q&A in the end.
8 Likes
@avehtari do you mind if I change the title of this to “StanCon 2024 tutorial thread”? And add my tutorial advert and a place for anyone to comment, ask questions, or post their tutorial for Stancon?
Hierarchical Models Tutorial at StanCon 2024
-
What: A 3 hour course on hierarchical models.
-
Who: The instructor is Sean Pinkney
-
Agenda: Background on hierarchical models, partial pooling and reparameterizations, and examples with meta-analysis, car insurance, and advertising.
-
New stuff for people who think they know hierarchical models: A derivation of a prior for partially centered parameters (i.e. where the weight is between 0 and 1), location-scale meta-analysis that shows the power of Bayesian models against frequentist, a data driven prior for marginalizing out id-level hierarchy when you want Bayes to go vroom.
-
Pre-reqs: Basic familiarity with Stan, Stan setup on computer (e.g. cmdstanr/cmdstanpy), and basic statistics.
Repo GitHub - spinkney/hierarchical_model_tutorial: Hierarchical Models Tutorial StanCon 2024. All code and slides will go there before the tutorial on Friday, Sept. 13.
3 Likes