TFPCA-Tutorial

Overview

The TFPCA-Tutorial serves as a companion to a tutorial paper introducing researchers to the time-frequency principal components analysis (TF-PCA) technique for EEG data. The tutorial is designed to help researchers run TF-PCA analyses using the psychophysiology toolbox (ptb). Moreover, this tutorial can serve as a jumping off point for more advanced analyses leveraging Cohen's class reduced interference distribution (RID) and/or TF-PCA on one's own data. The tutorial consists of template scripts to demonstrate how to:

1. Convert one's data into the format required by the psychophysiology toolbox (ptb);
2. Use the ptb to compute TF representations of both average (phase-locked) and total (phase and non-phase-locked) power via Cohen's class RID;
3. Compute tf-pca on average power and then copy pc weights over to total power by using the ptb;
4. Export variables for statistical analyses and conduct basic plotting.

Additionally, the TFPCA-Tutorial includes:

1. Example data (ERP CORE ERN) that was used with template scripts (https://osf.io/q6gwp/);
2. A copy of the psychophysiology toolbox 1.0.0 (http://www.ccnlab.umd.edu/Psychophysiology_Toolbox/);
3. A copy of eeglab2021.0 (https://sccn.ucsd.edu/eeglab/downloadtoolbox.php).

Currently, the TFPCA-Tutorial relies on MATLAB-based programming, and thus, requires that users have a valid MATLAB license to run the tutorial. Assuming a valid MATLAB license and install, the tutorial contains all additional toolboxes and scripts needed to run the analyses described (For tftb toolbox, please email Edward Bernat ([email protected]) with a name and an email address per the agreement with UMICH). Note that while the TFPCA-Tutorial currently relies on MATLAB, we plan to update the tutorial to remove the MATLAB requirement, either through an Octave or Python port of the code. Additionally, we have plans to update the tutorial to accept BIDS data and run within a fully containerized environment (Docker/Singularity). If you are interested in contributing to future developments for the TFPCA tutorial please contact us.

Quick Start

Before running the PTB, there are several steps to set everything up.

1. Have perl installed. Perl is generally installed with Linux and Mac OS. Type perl -v on a command line to determine the version of Perl installed. For Windows, perl is needed to be downloaded and installed (https://www.perl.org/get.html);
2. Have a valid MATLAB (The MathWorks, Natick, MA) license and have MATLAB installed;
3. Git clone this repository (or download directly);
4. Navigate to the analysis_template folder and run startup.m.

You should see messages indicating that the necessary toolboxes have been found :

Psychophysiology Toolbox veryfying and adding paths ... 

Verifying core toolbox paths ... 

FOUND: psychophys_components_path:      /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/components
FOUND: psychophys_dataproc_path:        /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/data_processing
FOUND: psychophys_dataimport_path:      /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/data_import
FOUND: psychophys_general_path:         /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/general
FOUND: psychophys_gui_path:             /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/gui

Verifying directories for scripts, cache, and output ... 

FOUND: dir_data_cache:                  ./data_cache
FOUND: dir_output_plots:                ./output_plots
FOUND: dir_output_data:                 ./output_data
FOUND: dir_scripts:                     ./scripts

Verifying paths for external bundled scripts ... 

FOUND: eeglab_path:                     /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/bundled_external_software/eeglab
FOUND: epsmerge_path:                   /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/bundled_external_software/epsmerge
FOUND: PCA_Toolbox_path:                /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/bundled_external_software/PCA_Toolbox

Looking for supported time-frequency toolboxes ... 

FOUND: DiscreteTFD Toolbox (GPL) - dtfd_toolbox_path:           /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/bundled_external_software/dtfd
FOUND: Rihaczek RID (GPL) - rid_rihaczek_toolbox_path:           /…/tfpca-tutorial/psychophysiology_toolbox-1.0.0/bundled_external_software/rid_rihaczek_PC
FOUND: Time-Frequency Toolbox (Q.S.) - tftb_toolbox_path:        /.../tfpca-tutorial/psychophysiology_toolbox-1.0.0/../tftb
FOUND: Matlab Wavelet Toolbox (Mathworks Inc.)

Psychophysiology Toolbox completed verifying and adding paths.

Tfpca-tutorial tests were performed using Matlab R2021a and macOS Big Sur (11.4). However, the tutorial should likely work with most other environemnts, on both Windows and Linux/Unix, but have not been explicity tested and may require minor modifications. If you run into issues with set-up and running of the tutorial, please post an issue.

Directory Structure & Scripts Descriptions

Below, we provide an overview of the contents of the TFPCA-Tutorial repository. Please note that PTB expects a particular directory structure, which is reflected in the directory structure of this repository. In other words, any changes in the directory structure of this repository may lead to errors. Of course, advanced users can modfiy the directory structure as needed, but it is not reccomended that beginners attempt to alter the directory structure.

|——psychophysiology_toolbox-1.0.0

Psychophysiology Tool Box (PTB) is a scripting-based Matlab toolbox developed by Edward Bernat and colleagues. PTB allows one to compute ERPs and TF representations (using Cohen's Class RID and RID-Rihaczek) and to decompose TF-PCA solutions on these data (http://www.ccnlab.umd.edu/Psychophysiology_Toolbox/). The PTB expects that data is formated in a particular manner and that a particular directory structure is used. Additionally, PTB expects that the user interacts with the toolbox by editing a series of scripts that must then be run from the appropriate working directory.

|——tftb

The Time-Frequency Toolbox (TFTB) is a collection of about 100 scripts for GNU Octave and Matlab developed for the analysis of non-stationary signals using time-frequency distributions (http://tftb.nongnu.org). A subset of these scripts are called by PTB, and thus, are needed to run the tutorial.

|——eeglab2021.0

EEGLAB is an interactive Matlab toolbox for processing continuous and event-related EEG, MEG and other electrophysiological data (https://sccn.ucsd.edu/eeglab). Note, to work with the example data (ERP CORE ERN), the erplab plugin (https://erpinfo.org/erplab) has been included. A copy of the EEGLAB toolbox (and the erplab plugin) are not neccesary to use the PTB, but are neccesary to run the provided example scripts for converting one's data into the format needed for PTB, as well as for the example scripts for plotting data after experting it from PTB.

|——eeglab_data

This folder is populated with a slightly modified version of the ERP CORE ERN. These data are used in the TFPCA tutorial.

The ERP CORE (https://doi.org/10.18115/D5JW4R) is a freely available online resource consisting of optimized paradigms, experiment control scripts, example data from 40 neurotypical adults, data processing pipelines and analysis scripts, and a broad set of results for 7 widely used ERP components: N170, mismatch negativity (MMN), N2pc, N400, P3, lateralized readiness potential (LRP), and error-related negativity (ERN). Only the ERN data are used for this tutorial.

Please note that the ERP CORE data available online (https://doi.org/10.18115/D5JW4R) were modified slightly to optimize for subsequent TF decompositions. The modified version of these data are included in the eeglab_data folder. The following steps have been taken to modify the data (in sequence):

1. Downloaded all 40 participants’ ERP CORE ERN data after processing step #4 (artifact-removed), which can be found here: (https://osf.io/ryk5u/);
2. Downloaded script #5 (https://osf.io/4whf6/);
3. Edited script #5 on lines 52 & 53 - changing the segmentation delimiters from [-600.0 400.0] to [-1000.0 2000.0] (See Script5_Elist_Bin_Epoch.m).
4. Ran the edited script #5 on the data downloaded in step #1 above.
5. Ran script #6 (which can be downloaded at https://osf.io/f3m7s/) on the data outputs from step #5 above. Note that script #6 was not edited following downloading from https://osf.io/f3m7s/. This script performs a final artifact rejection step that must be re-run since we modifed the epochs produced by the step #5 script (see Script6_Artifact_Rejection.m).

The steps listed above need not be completed by the user, as the modified data produced by these steps is already populated in the 'eeglab_data' folder. The modified data use the following naming convention: xx_ERN_shifted_ds_reref_ucbip_hpfilt_ica_corr_cbip_elist_bins_epoch_interp_ar.set.

|——ptb_data

This folder is populated with the ERP CORE ERN data that has already been converted to the format required by PTB. Therefore, the user does not need to convert the data found in the eeglab_data folder. However, the user is nonetheless provided with the data in the eeglab_data folder, and the script used to convert this data into the PTB format (template_eeg2ptb_erplab.m found in the scripts folder) in order to illustrate how to convert data from the EEGLAB/ERPLAB format into the PTB format.

|——analysis_template

This folder contains all scripts, files, and data that the user will interact with to run the tutorial. Additionally, this folder is pre-populated with the outputs resulting from running the included scripts. As described below, the analysis_template folder contains subfolders for example PTB scripts (ptb_scripts), example scripts for converting data to the PTB format, as well as for exporting and plotting the data outside of PTB (scripts), and several subfolders for various outputs created by running the tutorial scripts (data_cache, output_data, output_plots, exported_data). This folder also contains the startup.m script and the erp_core_35_locs.ced file.

|——————startup.m

This setup script adds all needed MATLAB paths (eeglab, ptb toolbox, tftb toolbox and other template scripts and output paths) and should be run first to set up the environment for the TFPCA-Tutorial.

|——————erp_core_35_locs.ced

This file contains the electrode locations file for the ERP CORE ERN example data included with the tutorial.

|——————ptb_scripts

This folder contains the PTB template scripts that can be edited and then executed in order to compute ERPs, TF representations, and decompose TF-PCA solutions.

|————————————run_Flanker_resp_ISFA_base_averages.m

|————————————Flanker_resp_ISFA_base_loaddata.m

|————————————Flanker_resp_ISFA_base_loadvars.m

Together, these three scripts will produce an averaged ERP dataset. run_Flanker_resp_ISFA_base_averages is the run script that is executed by the user, whereas Flanker_resp_ISFA_base_loaddata and Flanker_resp_ISFA_base_loadvars are scripts that act as parameter files are called by the run_Flanker_resp_ISFA_base_averages run script to load necessary parameters. Specifically, Flanker_resp_ISFA_base_loaddata designates basic information about where to find and how to process individual-subject data/files (file list and locations, time-domain baseline period, etc). Flanker_resp_ISFA_base_loadvars sets up several additional parameters, including "catcodes" (category codes; i.e. the conditions of interest), parameters for subsampling), electrode location files (.ced), output plot parameters (i.e. which electrode to plot) etc.. Furthermore, run_Flanker_resp_ISFA_base_loaddata calls load_Flanker_resp_EEG_subnames.m to loop over the data folder (../ptb_data) in order to get the list of each subject/data to be included in the analysis (each subject's data/file name will be used as the name for this subject). In most cases, load_Flanker_resp_EEG_subnames.m does not need to be edited as long as the data folder (../ptb_data) only includes subject files in the .mat PTB format.

In sum, run_Flanker_resp_ISFA_base_averages is the run script that is executed to produce an averaged ERP dataset, which is stored in 'data_cache'. Before running Flanker_resp_ISFA_base_averages, parameters should be edited in Flanker_resp_ISFA_base_loaddata and Flanker_resp_ISFA_base_loadvars.

Main outputs:

• data_cache:
  • Flanker_resp_ISFA_base_averages_128.mat - Averaged ERP dataset;
  • Flanker_resp_ISFA_base_averages_subsampling.mat - Subsampling dataset.
• output_plots :
  • Flanker_resp_ISFA_base_averages-win-rs128-StandardComps-plot_components - grand averages and components of interest;
  • Flanker_resp_ISFA_base_averages-win-rs128-StandardComps-plot_topo - mean and peak topographic maps of grand averages and components of interest;
  • Flanker_resp_ISFA_base_averages-win-rs128-StandardComps-plots_Merge_basic - a merged plot for two plots above.

|————————————run_Flanker_resp_AVGS_AMPL_theta_pcatfd.m

|————————————Flanker_resp_AVGS_AMPL_theta_DatasetDef.m

|————————————Flanker_resp_comparisons.m

Together, these three scripts compute "average power" (phase-locked power) TF representations and decompose TF-PCA solutions for these TF representations. run_Flanker_resp_AVGS_AMPL_theta_pcatfd is the run script that is executed by the user, whereasFlanker_resp_AVGS_AMPL_theta_DatasetDef and 'Flanker_resp_comparisons' are scripts that act as parameter files called by the run_Flanker_resp_AVGS_AMPL_theta_pcatfd run script to load necessary parameters. Specifically, Flanker_resp_AVGS_AMPL_theta_DatasetDef sets up various parameters, including the electrode_locations file to use, the TF transformation method to use, and the dataset name that will be used in all related outputs to 'data_cache', 'output_data', and 'output_plots'. Furthermore, Flanker_resp_AVGS_AMPL_theta_DatasetDef calls an additional script to perform filtering of the data (preproc_filter). Flanker_resp_comparisons defines the parameters associated with plotting and statistical comparisons between condtiions (catcodes).

In sum, run_Flanker_resp_AVGS_AMPL_theta_pcatfd is the run script executed by the user in order to compute average (phase-locked) power and decompose TF-PCA solutions. Before running Flanker_resp_AVGS_AMPL_theta_pcatfd, parameters should be edited in Flanker_resp_AVGS_AMPL_theta_DatasetDef, preproc_filter, and Flanker_resp_comparisons.

Main outputs:

• data_cache :
  • Flanker_resp_AVGS_AMPL_theta_32_t32f32.mat - the average power;
• output_data :
  • Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-PCs.mat - pc weights (based on the average power).
• output_plots :
  • Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_components.eps - time-domain grand averages across all conditions, time-frequency domain grand averages across all conditions (the average power) and pc weights (based on the average power);
  • Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_scree.eps - scree plot of variance being explained;
  • Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_topo.eps - topographical plots of mean pc weights, peak pc weights (based on the average power);
  • Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plots_Merge_long-resp_comparisons.eps - is the main plot that integrates all the plots genereated. Besides three plots described above, it further plots : 1) time-domain comparison(s), TF domain difference(s) and, pc-weighted TF difference(s) between conditions; 2) topographical plots of mean and peak pc-weighted TF difference(s) between conditions; 3) statistical significance of mean and peak pc-weighted TF difference(s) between conditions.

|————————————Flanker_resp_ISFA_AMPL_theta_pcatfd.m

|————————————Flanker_resp_ISFA_AMPL_theta_DatasetDef.m

|————————————Flanker_resp_comparisons.m

Together, these three scripts compute "total power" (phase-locked and non-phase-locked power) TF representations and decompose TF-PCA solutions for these TF representations. run_Flanker_resp_ISFA_AMPL_theta_pcatfd is the run script that is executed by the user, whereasFlanker_resp_ISFA_AMPL_theta_DatasetDef and Flanker_resp_comparisons are scripts that act as parameter files called by the run_Flanker_resp_ISFA_AMPL_theta_pcatfd run script to load necessary parameters. Specifically, Flanker_resp_ISFA_AMPL_theta_DatasetDef sets up various parameters, including information about how to find and process individual-subject data, parameters for subsampling, the electrode locations file to use, and the dataset name that will be used in all related outputs to data_cache, output_data, and output_plots. Furthermore, Flanker_resp_ISFA_AMPL_theta_DatasetDef calls an additional script to perform filtering of the data (preproc_filter). Flanker_resp_comparisons defines the parameters associated with plotting and statistical comparisons between condtiions (catcodes).

In sum, run_Flanker_resp_ISFA_AMPL_theta_pcatfd is the run script executed by the user in order to compute total (phase-locked and non-phase-locked) power and decompose TF-PCA solutions. Before running Flanker_resp_ISFA_AMPL_theta_pcatfd, parameters should be edited in Flanker_resp_ISFA_AMPL_theta_DatasetDef, preproc_filter, and Flanker_resp_comparisons.

Main output:

• data_cache :
  • Flanker_resp_ISFA_AMPL_theta__32_t32f32.mat - the total power;
• output_data :
  • Flanker_resp_ISFA_AMPL_theta_-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-PCs.mat - pc weights (based on the total power).
• output_plots :
  • Flanker_resp_ISFA_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_components.eps - time-domain grand averages across all conditions, time-frequency domain grand averages across all conditions (the total power) and pc weights (based on the total power);
  • Flanker_resp_ISFA_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_scree.eps - scree plot of variance being explained;
  • Flanker_resp_ISFA_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plot_topo.eps - topographical plots of mean pc weights, peak pc weights (based on the total power);
  • Flanker_resp_ISFA_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-plots_Merge_long-resp_comparisons.eps - is the main plot that integrates all the plots genereated. Besides three plots described above, it further plots : 1) time-domain comparison(s), TF domain difference(s) and, pc-weighted TF difference(s) between conditions; 2) topographical plots of mean and peak pc-weighted TF difference(s) between conditions; 3) statistical significance of mean and peak pc-weighted TF difference(s) between conditions.

|————————————cp_avg_power_pcs.m

As described in the companion paper and tutorial walk-through, it is often useful to copy the PC weights derived from applying TF-PCA to an average (phase-locked) power TF representation and then apply these weights to a TF representation of total (phase-locked and non-phase-locked) power. PTB does not currently have built-in functionality to facilitate copying of PC weights derived from one TF representation to another. However, this can be achieved realtively easily by first running the run_Flanker_resp_AVGS_AMPL_theta_pcatfd script, and then running the run_Flanker_resp_ISFA_AMPL_theta_pcatfd script. Next, within the output_data folder, a copy of the Flanker_resp_AVGS_AMPL_theta-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-PCs.mat file needs to be made and subsequently renamed with the same name of the file generated from running the run_Flanker_resp_ISFA_AMPL_theta_pcatfd run script: Flanker_resp_ISFA_AMPL_theta_-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-PCs.mat. There cannot be two files with the same name, within the same folder, and thus, the origional Flanker_resp_ISFA_AMPL_theta_-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1-PCs.mat file must be deleted (i.e. it is being replaced by the PC weights generated by the run_Flanker_resp_AVGS_AMPL_theta_pcatfd run script). Additionally, the Flanker_resp_ISFA_AMPL_theta_-pcatfd-rs32-t32s-16e16-f32s7e19-fqA1-DMXacov-ROTvmx-fac1.log must also be deleted, and corresponding files in output_plots must also be deleted. At this point, one can re-run the run_Flanker_resp_ISFA_AMPL_theta_pcatfd run script and the copied PC weights will be used for the generated outputs in output_data and output_plots.

While the manual series of steps described above to copy/rename/delete files will work, this approach is tedious, and most importantly, error-prone. Thus, the user can instead use the cp_avg_power_pcs.m template script to perform these manual steps automatically. After running cp_avg_power_pcs.m, Flanker_resp_ISFA_AMPL_theta_pcatfd can be run again to generate outputs to output_data and output_plots using the copied PC weights.

|——————data_cache

|——————output_data

|——————output_plots

These three folders are populated by the outputs generated by the run scripts located in ptb_scripts (and described above). Specifically, data_cache stores computed ERP and TF data prior to any region-of-interest (ROI) analysis or PC-weighting. output_data stores computed PC weights for any extracted TF-PCA solutions, as well as PC-weighted TF representations, and, if caculated, ROI analyses are also stored here. output_plots stores various plots (waveforms, topographic (topo) maps, statistical topo maps, etc.) generated while running the scripts located in ptb_scripts. Note that the plots generated by PTB by default and stored in 'ptb_scripts' should not be used for publication, and we recommend that users utilize the plotting scripts found in the scripts folder in order to generate publication-ready plots.

|——————scripts

This folder contains non-PTB scripts, including template scripts for converting EEGLAB/ERPLAB formatted data to the PTB (.mat) format, template scripts for converting the data produced by PTB into a format that is easier to manipulate for plots and extraction of values for statistical analyses, and template scripts for plotting data outside of PTB. This folder also contains the cp_avg_power_pcs.m for copying PC weights within PTB, as described above.

|————————————template_eeg2ptb_erplab.m

This template script converts EEGLAB/ERPLAB formatted data to the PTB (.mat) format and stores it in ../ptb_scripts

|————————————template_ptb_cache_out.m

This template script includes example code to convert either average (phase-locked) power or total (phase-locked and non-phase-locked) power data (located in ../data_cache) from the PTB format to a format that is easier to plot/analyze outside of PTB. By defualt, this template scripts is set up to export the converted data to a file called AvgPower_resp_TFD.mat (for the exported average power data) and TotalPower_resp_TFD.mat (for the exported total power data) to a folder titled, ../exported_data.

|————————————template_ptb_pc_out.m

This template script includes code to use the PC weights generated by PTB to weight the reformatted TF representations produced by the template_ptb_cache_out.m script (stored in ../exported_data) and produce PC-weighted TF representations for average (phase-locked) and total (phase-locked and non-phase-locked) power, which are saved as Theta_AvgPower_resp_TFD_pcWeighted.mat and Theta_TotalPower_resp_TFD_pcWeighted.mat, respectively. These PC-weighted TF surfaces are generated in a format that is easy to plot and analyze outside of the PTB and are saved in the in the ../exported_data folder.

|————————————template_dbpower.m

This template script conducts dB power conversion to the average power and the total power respectively. The generated data (AvgPower_resp_TFD_baseRemoved.mat for the average power with dB power conversion and TotalPower_resp_TFD_baseRemoved.mat for the total power with dB power conversion) was stored in ../exported_data.

|————————————template_plots.m

This template script plots the results from template_ptb_cache_out.m, emplate_ptb_pc_out.m and template_dbpower.m. Specifically, it plots the total power with dB power conversion, pc-weighted total power, topographical plots of pc-weighted total power, and the pc weight itself.

|——————exported_data

This folder is populated with data exported from PTB and stored in a format that is easy to plot/analyze outside of PTB. The data located here is generated by running the aforementioned scripts located in the scripts folder.

|——new_analysis_template

This folder is a copy of the analysis-template folder, but with all output files in the data_cache, output_data, output_plots, and exported_data folders removed. The new_analysis_template should be used when following along with the tutorial and running the scripts without needing to delete the example outputs.

Glossary

Cohen's class reduced interference distribution (RID) - A time-frequency transformation method yielding improved time-frequency resolution without requiring a priori parameterization for a restricted frequency band.

Time-frequency Principle Components Analysis (TF-PCA) - A data reduction technique that allows for isolating distinct phenomena within the TF representation. TF-PCA involves application of principal component analysis (PCA) to the time-frequency representation after first converting each TF-representation into a long vector by conactenating frequency bins across time.

Average power - A TF representation computed from time domain data that has already been averaged across trials of interest. Thus, average power contains primarily phase-locked power data.

Total power - A TF representation that is computed for each trial, and then averaged across trials only after TF decomposition. Thus, total power contains both phase-locked and non-phase-locked power data.

PTB Format - Data and index variables are stored together in a structured variable: cnt (continuous), erp (epoched), components (derived measures from erp variable). The PTB toolbox operates as a flat-file database (a 'univariate' data setup). A main 2-d data matrix (trials by waveforms) is indexed by vectors the same length as trials. Further details of this structure are available in the documentation directories (psychophysiology_toolbox-1.0.0/documentation/data_import/README_dataset-structure.txt)

How to Cite

If you use these resources, please cite all three of the following:

1. The original TF-PCA methods paper: Bernat, E.M., Williams, W.J., Gehring, W.J., 2005. Decomposing ERP time–frequency energy using PCA. Clin. Neurophysiol. 116, 1314–1334.
2. This github repository: https://github.com/NDCLab/tfpca-tutorial
3. The companion tutorial article: Buzzell, G.A., Niu, Y., Aviyente, S., Bernat, E.M., under review. A Practical Introduction to Time-Frequency Principal Components Analysis (TF-PCA) of EEG Data.

Contact Us

For further information, questions, or feedback, please contact:

George A. Buzzell, Ph.D - [email protected]

Neural Dynamics of Control Laboratory; Department of Psychology and the Center for Children and Families (CCF); Florida International University, Miami, Florida

Yanbin Niu, MA - [email protected]

Neural Dynamics of Control Laboratory; Department of Psychology and the Center for Children and Families (CCF); Florida International University, Miami, Florida

Edward Bernat, Ph.D. - [email protected]

Department of Psychology, University of Maryland, College Park