search

francescomantegna / mlm_music

mixed linear model for a tonal prediction EEG experiment
GNU General Public License v3.0
0 stars 0 forks source link

No Participant-Level Exclusions Based on Trial-Level Rejections #5

Closed palday closed 4 years ago

palday commented 4 years ago

Verify that even participants with large number of trial-level rejections are in the dataset. I've been harping about this for a while now, but Ben Bolker said it succinctly recently:

This is a very superficial comment that doesn't really engage with the details of your struggle, but in general it makes me sad when people drop observations, especially from mixed models, and especially in a Bayesian context. In principle (I know it doesn't always work out this way in practice), observations that are only weakly informative are still giving you some information ... is there any way you can solve the problem by setting stronger (i.e. more strongly regularizing) priors ?

@francescomantegna Can you verify that this is the case?

francescomantegna commented 4 years ago

Yes, we did not exclude any participant because of large trial-level rejections.

palday commented 4 years ago

Excellent.