New OxACCLab paper reviewing studies of confidence in an important applied setting -- medical decision making. First paper from Sriraj Aiyer's fantastic PhD! In collaboration with Helen Higham @HelenEHigham https://t.co/AwBBcQcG03
Out today! EEG signatures of automatic and domain-general representations of confidence:
https://t.co/SSbAMgwuiv
Project led by the fantastic @matt_j_davidson
New ACCLab paper from Michael Ben Yehuda's PhD: behaviour, EEG and Bayesian modelling showing how confidence in learning governs people's attention to feedback they receive
https://t.co/SsWzBJCkee
@lakens (For the specific query, the answer is that the pre-reg I linked to earlier was Exp 3 of that paper. The quadratic trend was our original prediction for Exp 1, which relates to the results in your screenshot -- the Exp 1 pre-reg is here: https://t.co/PjADviarLV)
@lakens I guess I'm asking a general question about the aspredicted philosophy (as I understand it), that there's value in the basic case where you pre-register a specific analysis and report exactly that analysis (especially where, in principle, the analysis could be done many ways).
@lakens I'm not sure why it was a waste of time. We pre-registered the key analysis, which critically (for me) was an analysis we could in principle have run in many different ways to test our hypothesis, and that was the analysis we ended up reporting https://t.co/q9hJcW4DmD
@lakens About alternative analysis plans (e.g., if test assumptions are not met), we sometimes do this. But even if not, the pre-reg has value because we can specify clearly where we've departed from our original plan and need to justify why.
@lakens There are lots of things that need to be in the final paper (such as detailed methods) that in principle could be included in a pre-reg but which I prefer to keep out, so that the pre-reg focuses on restricting researcher degrees of freedom in data collection and analysis.
@lakens@rmwillen I had the same type of question @rmwillen and see similarities in our pre-reg approach (with similar deficiencies in the eyes of @lakens!)
@lakens About sample sizes, I'm not sure a pre-reg is where that detail/discussion fits. We use pre-reg to pre-determine our analyses (however inadequately...) not as a cure-all for experimental design
@lakens Thanks for prompting the discussion. Our lab typically uses aspredicted, not because it's short but because it focuses on analysis (rather than methods). I think the nudge in their question "Specify exactly which analyses you will conduct..." is exactly right for a pre-reg.
New ACCLab paper -- Bayesian Confidence in Optimal Decisions. By the fantastic Josh Calder-Travis in collaboration with Rafal Bogacz and @LucieCharlesCog
https://t.co/zFCLMs4PLs