To do
=====
[ ] BCI utility & another metric i don't recall
[ ] DIC or BF / TEST
[ ] Posterior predictive check, bayesian p-value
- check literature for important statistics
- P300 feature extractions
https://projecteuclid.org/journals/annals-of-statistics/volume-22/issue-3/Posterior-Predictive-p-Values/10.1214/aos/1176325622.full
https://pubmed.ncbi.nlm.nih.gov/19174332/
www.stat.columbia.edu/~gelman/research/published/A6n41.pdf
[ ] pre-estimate smoothness to choose a parameter
[ ] EEGNet https://github.com/vlawhern/arl-eegmodels/blob/master/examples/ERP.py
Notes for next meeting
======================
Ran the experiment by reversing the order of the initialization
- Basically what I expected: the initialization dictates the result,
and the prior has almost no effect (some small shrinkage).
The posterior for the shrinkage factor is not longer decreasing.
Turns out a1 should be chosen to be 1 instead
Backlog
=======
Commands
========
ssh simon@192.168.0.208
scp <files> simon@simfont:/home/simon/Documents/BCI/experiments/test3/chains
# get from GL
scp simfont@greatlakes-xfer.arc-ts.umich.edu:./Documents/BCI/experiments/k114/chains/* ./experiments/k114/chains