Cognitive (neuro) science StanConnect 2021

Hi,
I would like to organize a StanConnect meeting so that we can hear about the projects of the cognitive (neuro) science community.

Are there people interested? Could people answer if they would like to attend or even to present? For people that want to present, it would be great to have a short (~250 words?) abstract, and we could vote to choose the speakers. I’m also open to other ideas, but the template of 3 short talks could work nicely.

Tentative date would be sometime between October-December.

Best,
Bruno

Update

The call for the abstract submission here is closed.

Please vote for the abstracts here.

19 Likes

Count me in! What’s the time frame of when it might happen?

Pertinent topics on which I’m already decently prepared to do a talk on:

  • psychometrics of hierarchical/multilevel models (inc. my recent explorations of computing icc/reliability as well as the limitations of the multivariate normal)
  • hierarchical/multi-level models for speeded choice data, with errors and RT location/scale modelled together for optimal information-sharing and mixture-modelling of fast and slow RT outliers

Less-ready but something that’s long been in my to-do list and recently worked on a bit:

  • GPs for neuroimaging (solves the “cluster-failure” issue that got press a few years ago)
5 Likes

ah good point. I would say the last part of the year? Between October and December. I’m adding that to the original post.

I’d be interested. Contribution could be sth like “case study: inferring meaning representations from language use” and the difficulties joys with these kinds of (largely discrete) models in Stan?

4 Likes

I’d be interested too!

I’m using reinforcement-learning drift-diffusion models enriched by trial-by-trial eye-tracking and MEG measures (similar to the HDDM toolbox by Wiecki, Sofer & Frank that uses PyMC).
Idea is to in the end also link the drift process itself to neural measures (some ideas, but nothing conclusive yet).

Limitations:

  • MEG data collection is about to start this month,
  • my PhD contract might be over in November this year (possibility of extension),
    so
  • can’t commit 100% yet
  • can’t show many results as of yet… :-/
4 Likes

This sounds very interesting. At the very least, I’m interested in attending.

I could contribute… My group has been working on the application of location-scale models to ERPs across psychopathology groups. We are… mid project, but I hope to wrap it up in the next few months (once the semester is over).

I’m not sure what the audience is at StanConnect. I imagine it’s full of people that all understand bayesian mlm better than I, so it certainly sounds intimidating :) Are there recordings of previous talks somewhere? Nonetheless, I would be interested in applied talks related to cognitive neuroscience.

3 Likes

This will be the first stanconnect and the first time there is a session for cognitive and neuro science. You can check previous talks in Stan - Events. I presented twice but I felt quite alone with my topic :) hence this.

I guess the audience is people using Stan that want to know what neat models and tricks other people are using to deal with experimental data.

I really liked the notebooks in the StanCons I attended, a small rmarkdown document with all the code. One can learn a lot from what other people do, and that level of technical details is usually absent in papers. So I would like to have that as well with the presentations.

3 Likes

I would love to attend this!

2 Likes

I’m interested in attending! I also use hierarchical Stan models (through the brms package in R) to model longitudinal changes in functional brain development if you are interested in developmental cognitive neuroscience work.

Topics which the presentation could hit on include:

  • Parametrizing within-participant versus between-participant terms in hierarchical regression models
  • Doing ‘multiverse’ analyses and constructing specification curves with Stan-based models
  • Using t-distributed (vs. gaussian) likelihood functions to increase regression model robustness to outliers
4 Likes

I’d love to attend!

1 Like

Thanks for the responses, based on the discourse thread and twitter, it seems there are quite a few people interested.
I now realize that there is a deadline quite soon for the proposal: April 26, 2021 (New York).

I guess it makes sense to send a call or at least a link to this thread to some lists, right? That will for increase the diversity of speakers (outside Stan discourse and my twitter bubble).
Maybe Mathpsych and CMCL (Cognitive Modeling and Computational Linguistics) lists, any other ideas? It has to be in the next couple of days! Then at least 3 people need to commit, I understand that at least @mike-lawrence, @michael-franke, and @paul_a_bloom can commit, and I have another committed person by email.

After getting the final potential topics, we’ll need to either vote (for three) or to allow for more talks but shorter. I think that 10 minutes talks are also a possibility, but I’d like to hear opinions.

Another issue is the notebooks, does it make sense to have short notebooks for every talk, similar to what is was done for Stancon? Or is it an overkill?

Edit: A notebook is just an rmarkdown or jupyter document with the analysis.

I like both the notebooks idea as well as shorter talks to enable more of them, maybe followed by something akin to the post-colloquium casual discussion time that academic departments often have. I don’t know what platform would enable the latter remotely but I’d be surprised if it weren’t implemented somewhere.

2 Likes

yeah, the two templates include a “Zoom Networking/Social session. Each of the three speakers will have their own breakout room. Audience members can go in and out of the rooms to chat/mingle.”

1 Like

I have nothing substantive to add, but I like the notebook idea and the breakout groups. Hopefully the talks can also be recorded for posterity.

Thanks for spearheading this @bnicenboim!

1 Like

Regarding dates, is there a preference for during the week or a weekend?
End of November would be the most convenient for me, is there some important conference that I should avoid to overlap?

For those with childcare needs weekends can sometimes be more challenging that weekdays (but finding time during weekdays can also be challenging).

3 Likes

haha, I wholeheartedly agree that both weekdays and weekends are challenging with childcare needs. But I was also thinking about time zones. I guess people can be more flexible on a weekend?

ok, let’s say that the session will be on Friday November 26th, and we’ll decide on the time depending on the location of the majority of the speakers.
If no one complains until evening Europe, I’ll make a call, and I’ll set up some other page for the abstracts and the voting.

The idea would be to have the short abstracts and a commitment to present a short talk and provide a notebook by Wednesday 21st April. The notebooks should be ready by early November and could be hosted either in the speakers’ webpage or in cognitive-science-stan.github.io

2 Likes

I would also like to attend. Im currently working on the implementation of the memory measurement model (M3; Oberauer & Lewandowski, 2019) framework in stan for easy application in cognitive research and benchmarking the model. It models different working memory tasks, e.g. complex span.

1 Like

I would definitively attend to a session on Cognitive and Neurosciences at the StanConnect

1 Like