Quoting @gangchen6.bsky.social:
"Next up: bringing this to everyday analysis.

AFNI’s new program SIMBA is in development and aims to make full whole-brain voxel-level hierarchical modeling accessible to users, hopefully within the next few months."

Maybe he'll make for us a great Xmas present! 😂

If your ultimate inference target is the group level, then what matters is the joint contribution of trial number per condition and participant sample size, not either one in isolation. We explored this point in detail here:
www.sciencedirect.com/science/arti...

@pessoabrain.bsky.social - that’s something that you’re very familiar with as well. Thoughts about the minimum number of trials per condition to run bayesian multilevel modeling?

As I said to Vinny in the past, I think this paper is simply incredible.
There’s one thing that I’m curious about these analyses: based on your experience, how many trials per condition do you think would be a decent amount to good estimates?

What could we do to stop this plague? I keep telling students that they are doing what they are for their own good, not for grades..but many just don’t seem to care. Higher performance with minimum effort, that’s all that seem to matter

I might be wrong on this, but sometimes I have more of the “wow factor” by reading papers from the 90s than I have when reading the lasted flashy paper on nature something.
It might just be me..but I have the feeling that we are just accumulating papers, not knowledge.

But how about..slowing down? The publication rate is too high, because the system we built is meant to foster careers and not good science. And all this is doing is making big publishing groups reach at the expenses of researchers work.

but that's the plot twist: they are pushing us to fight to keep the DEI 😂

@davidbaranger.bsky.social - I’m genuinely curious about your view here. I know your work on reward processing and reliability and see it as a reference point in the field, so I’d be really interested in any nuances you’ve picked up after working on these questions.

3/3
Knocking down ERα in midbrain DA neurons blunts sensitivity to reward context without changing thirst. Nice link between estrous cycle, dopamine RPEs and reinforcement learning, with obvious implications for sex differences & menstrual-cycle effects in psychiatry.

2/3
Using GRAB-DA photometry + modeling, NAcc dopamine encodes reward prediction errors, and high estradiol boosts especially large positive RPEs. Proteomics points to a mechanism: reduced DAT/SERT expression → slower reuptake → bigger phasic DA signals.

And the next step? Full voxel-level modeling.

Recent numerical advances cracked the scalability barrier. Voxel-level hierarchical modeling is now feasible, revealing just how punishing traditional multiple-comparison adjustments really are.
arxiv.org/abs/2511.12825

3/
fMRI: despite inhibiting vmPFC, cTBS amplified reward-prediction-error BOLD in vmPFC, mediodorsal thalamus, and dorsal striatum.
Authors interpret this as a shift from fast, vmPFC-driven Pavlovian invigoration toward slower, more uncertain thalamo-striatal learning. #dopamine #fMRI

2/
Design & behavior: single-blind vmPFC cTBS vs sham before a motivational Go/NoGo task in the scanner.
cTBS →
- fewer Go responses
- slower RTs
- RL modelling: selective drop in positive learning rate (gains learned more slowly), trend toward reduced Pavlovian bias.

That’s an incredibly accurate analogy