Hey folks, I’m helping supervise @kn2465’s work on the Benchmarking GSOC project and we’re looking for structurally-interesting models to include, ideally with some associated description in existing literature. I have seen lots of models in the cognitive science world that fit this description, but am admittedly not good at reading/keeping-track of literature, so I thought I’d crowdsource. Have any suggestions?
As an example, there’s a neat family of VonMises-Uniform mixture models being explored in the immediate-memory literature.
7 Likes
What does “structurally-interesting” mean?
I’d say models with more complex structure than simply a GLM/hierarchical-GLM.
We have been fitting non-linear hierarchical models of economic choice . We have a paper out using brms in intertemporal choice (https://elifesciences.org/articles/39656) with all the code and data shared on github (GitHub - erlichlab/delay3ways: Our study on bridging the gap between animal & human studies of delay-discounting). We are working on some papers of choice under uncertainty. Once they are ready, the code and data will be also shared.
2 Likes
Ah, perfect! This will make for a good addition for sure.
Just a head’s up: we made a very rookie mistake fitting our data as bernoulli instead of binomial. If you compute the counts ahead of time in R and fit as binomial it is much faster and gives the same results.
2 Likes
Cool, thanks for the heads-up. The binomial sufficient-stats trick is definitely on our list of performance tricks we’ll be using when we get to the stage of optimizing models.
1 Like