Adding hierarchical error bars to PSTHs

[alexpiet]

Uses https://github.com/AllenNeuralDynamics/aind_hierarchical_bootstrap to implement hierarchical bootstrapping to provide error bars and significance testing for PSTHs

example:

fig, ax, etrs = plot_fip_psth_compare_alignments(
    nwbs, 
    alignments, 
    channel_name, 
    error_type='hb_sem',
    hierarchical_params={'nboots':10000}
)

etrs is a dictionary that contains the time/values for each PSTH, as well as the bootstrapped values for each channel or alignment, and statistics dataframe for all comparisons performed.

Can add significant results to the plot with:

fip_psth_stats_plot(ax, etrs['stats'], color='k', threshold=0.05)

The README for the aind_hierarchical_bootstrap has a demonstration of why this approach is needed, as well as verification the implementation works: https://github.com/AllenNeuralDynamics/aind_hierarchical_bootstrap

This code ocean capsule has a demonstration of this toolset: https://codeocean.allenneuraldynamics.org/capsule/5461482/tree, /code/hierarchical_bootstrap_demonstration.ipynb

[rachelstephlee]

example capsule is private-- i can't view it!

[rachelstephlee]

can you add more what this stats_df will have?

also, note that the two plot_fip function will return this as part of etrs.

[alexpiet]

Added more documentation

[rachelstephlee]

I think I mostly get this. The code is really clean, the example capsule would be crucial.

So if I want to have hierarchically fitted results for per mouse, per session, I would have to use two levels in the hierarchical bootstrapping, levels=['subject_id', 'ses_idx'], giving me statistics of significance per session, per subject. Is that correct?

[rachelstephlee]

looks and works great!

can you put the example notebook in this capsule so we can have all sample code together?
https://codeocean.allenneuraldynamics.org/capsule/6488619/tree

I think you just need to cut down the number of nwbs so that it's just 10 nwbs to pull.