spec-state-trans: Spectral state transitions project

Dependencies:

• proektlab/Ethan_misc - clone next to this repository

  • Add Ethan_misc to the MATLAB path.
  • Edit Ethan_misc/prepSR.m to point to local folders, as described in the comments.

• proektlab/SpectralAnalysis

• proektlab/NonNegativeMatrixFactorization

Current analysis pipeline

See mt_new_2020_recs.m followed by full_analysis_byday.m for an example.

1. Compute CSDs for each region on each recording day using plot_csd.m

   • Output: <date>/csd_<region>.mat

2. Use CSDs to assign channel numbers to layers in the CSD matfiles using pick_channels_from_csd.m

   • Output: <date>/csd_<region>.mat

3. Find burst suppression using find_likely_bs.mat and mark any other artifacts manually in the artifacts option to multitaper_analysis

4. Do multitaper analyses of each recording's selected channels, with artifact filtering etc. using multitaper_analysis.m

   • Output: <date>/<time>/mt_res_layers.mat

5. Do NMF and extract classes and transitions for each channel (based on dominant component per time) using concat_and_nmf.m

   • Output: <date>/nmf_res.mat

6. Generate "null model" score data for each channel (shuffled using Markov chain - negative control for connectivity) using gen_null_model_data.m

   • Output: <date>/nmf_res.mat

7. Compute Euclidean distance between aligned NMF scores of each directed pair of channels (and plot results) with score_dist_analysis.m

   • Output: <date>/nmf_res.mat

8. Get normalized mutual information between discrete classes of each channel using class_mut_info.m from Ethan_misc (see full_analysis_byday.m)

9. Compute a measure of synchrony for each transition and compare to bootstrap with transitions shuffled using get_state_transitions.m and calc_transition_synchrony.m