Interest in participating

[behinger]

Hi! Very nice project you have been starting. I'd be interested in contributing.

Some of my potentially useful experience:

• hierarchical models (extensively via MixedModels.jl / LME4)
• okaish to solid STAN bayes knowledge especially with hierarchical models, but no super crazy models ;)
• multiple regression / GAMs / generalized models
• Bayesian "Cognitive" Modelling ala Wagenmaker et al. (not a lot of applied experience though)
• sequential sampling (nothing super serious / experienced, I'd like to learn more)
• Makie.jl plotting + some development (e.g. TopoPlots.jl)
• Julia development (e.g. our toolbox family: https://github.com/unfoldtoolbox/unfold.jl/)

Why I want to contribute?

• I want to dive closer to cognitive models, maybe even to something high-level like ACTR.jl at some point (>5 years)
• I think I have some unique intuitions on these topics, and many I have been explaining for years to students (www.benediktehinger.de has some of the explanations for eternity, e.g. on contrast coding)
• I like Julia and want to push it more :)
• I will likely give a course on these topics to undergrads at some point, so why not use this as lecture preparation

Some random thoughts:

• The topics you propose right now read similar to Farrell & Lewandowsky - Some years ago I gave a seminar on that book. I think computational modelling fits better than cognitive models, but maybe that's just because I never really understood why calling drift diffusion models cognitive models is accepted 🤷.
• It would be relatively easy for me to work on the chapter you call "predictors", which given the topic, works on multiple regression, GAMs, and the "generalized" aspect of multiple regression to other scales - I have dabbled into bayesian ordinal regression (Paul Bürkners work) but without serious application
• I'd be interested to be involved in the sequential sampling topics

Finally, I'm wondering what other collaborative books on similar topics are out there, and which we might be able to borrow/link ideas & in what way the idea is to create something more original. For example Vashisth has this book: https://vasishth.github.io/bayescogsci/book/ which is similar in ideas. There is the Bayesian Cognitive Modelling by Wagenmaker & Lee, the Farrell book, probably more if I search for them. Maybe the goal is to introduce a way to do these topics inside Julia with some more superficial descriptions of the concept and links to other books? A bit of clarification if the goal is set already, and what it then is would be great! I'm open for a lot - I will also see if I can convince some of my team to contribute ;-).

Disclaimer: As probably familiar to you, at this stage in my career there are a lot of interesting projects one wants to spend time in, but not enough time. I have to think how much time I can commit to this project.

[DominiqueMakowski]

Hi @behinger, I'm so happy to have you on board (as you might remember, I had Unfold on the radar since its creation ☺️ )

It would be relatively easy for me to work on the chapter you call "predictors"

This is definitely what came to my mind. You're probably the biggest GAM expert in Julia, so having your help for this part would be amazing!

I'd be interested to be involved in the sequential sampling

That's currently the main thing I'm planning to use Julia for. I think it will be made possible thanks to the incredible of @itsdfish on https://github.com/itsdfish/SequentialSamplingModels.jl. I would recommend you to follow that package, we have a lot of discussions there it's quite active; I learned a ton from @kiante-fernandez (who is making a talk on that at JuliaCon) and @itsdfish

I think computational modelling fits better than cognitive models

The terminology is not set in stone, so definitely open for discussion. I think "computational" is a bit overused though (computational neuroscience, computational psychiatry, computational ...) and has a lose definition, whereas "cognitive (statistical) models" means - in my mind - models 1) suited for cognitive/behavioural data and 2) especially models that can help making inferences on potential underlying processes (tlike DDMs & co)

https://vasishth.github.io/bayescogsci/book/

Yeah, this is a great one. Although I see it as probably too much in depth as compared to what I had envisioned for this.

I don't plan to have a lot of equations and in-depth discussions about stuff, I would just like to provide an overview to users of 1) why what they currently do is probably not optimal, 2) what they can do to make the most out of their data, and 3) how can they go about it to give it a go.
An in-depth intro to either Julia or Bayesian stats or Mixed models or Sequential Sampling Models or any of the particular topic is outside the scope (they would all deserve a separate book).

Here the goal is almost to present like an "overview of good recipes for common data in cog psychology & neuroscience".

I have to think how much time I can commit to this project.

Same 😅 let's take it slowly. At least, I know I can tag you in issues to ask questions on GAMs if need be ;)

[itsdfish]

Welcome @behinger! I am the primary maintainer of SequentialSamplingModels.jl, ACTRModels.jl and its companion package ACTRSimulators.jl . If you are interested in SSMs or ACT-R, one easy way to start contributing is to look through the code and add an API design page to the documentation. Documentation explaining the basic API design and how to create a new model/feature would be useful to users and contributers alike. Please feel free to open a discussion if you decide to contribute and have any questions.

[DominiqueMakowski]

For GAMs, I asked some time ago if GAMs.jl could be the package to do some of the heavy lifting to generate the stuff we need to fit them easily in Turing, but I'm not sure if there's a future to that