These files perform analysis on electrophysiology data taken while subjects perform the balloon analogue risk task (BART). Most data handling is done in Python, with some plotting and specialized analysis done in R. Data are stored in an HDF5 file.

• build_db.py defines a DataSets class that constructs the hdf5 database from Matlab and Plexon source files. Individual files for each task (in subdirectories) use this file in constructing databases.

• build_bart_db.py uses the DataSets class from build_db.py to construct the BART database.

• correlation_in_time.py uses local eigenvalue decomposition of the covariance matrix between channels to extract a time-varying measure of the effective dimensionality of the dynamical system underlying LFP measures.

• corrplot.py uses multidimensional scaling to plot the relationships among LFP channels as a function of time, using cross-channel correlation as the distance measure.

• prep_classifier_data.py collects multichannel field potentials for each dataset and processes them into a form suitable for fitting predictive models of behavior. This processing involves constructing a matrix of integrated power in each channel in specified frequency bands for each trial (to use as regressors). Data are written to csv files.

• plot_channel_traces.py constructs a peri-event line plot of all channels, averaged across trial.

• plot_LFP_channel_raster.py constructs a heatmap plot of peri-event LFP power for a single frequency band and channel. Each row in the plot is a trial.

• do_lfpglm_analysis.R calls the function in dolfpglm.R to run a glmnet regression on data for a specific dataset. Saves fit objects to disk.

• process_lfpglm_analysis.R loads the fit objects, calculates performance (captured in an ROC curve), and produces a plot (saved to disk). Details of analysis, data partitioning, and cross-validation are in run_lfp_glm.R. Uses miscellaneous functions defined in helpers.R and setup_env.R.

• Python: NumPy, SciPy, Pandas, h5py (for reading Matlab files), pytables (via Pandas), warnings, rpy2 (for calling R). In addition, makes extensive use of the physutils repository.

• R: glmnet, ggplot2, reshape.

• Task-specific code needs to import build_db.py, physutils.py, etc., so these need to be somewhere in PYTHONPATH.