EEG_RSA

Toolboxes for my RSA analysis, with associate classes and functions. There are two kind of files here: independent toolboxes as classes or standalone functions, and main functions that requires the toolboxes to function.

Author

Feng Cheng, Tufts University, Research Assistant at NeuroCognition of Language Lab

Environment File:

RSA_env.yml

Toolboxes:

variables_processing_alt.py

• Description A class for storing variables. It is recommended to store all variables together into one instance, since many functions operate on multiple variables simultanously.
• Status: In Progress. Lack comments; need to rearrange some functions for simpler interactions.
• Functionality (refer to individual functions for details)
  • Calculate Representational Disimilarity Matrix (RDM) using a customizable function
  • Manage the variables through applying general masking/re-ordering, extracting trials from specific conditions, and interpolation (a simple 1-dimesional nearest-neighbor algorithm is included). All of those management can be specified simultanously.
  • Export variables and associated RDMs
  • Prepare upper-triangular RDM for RSA.
  • Calculate correlation (Partial/Non-partial) between variables values and their RDMs
  • Impute missing values with iterative imputation (not tested yet)
  • Calculate cosine similarity from a given word embedding corpus

Single_Trial_RSA.py

• Description A class for performing representational similarity analysis (RSA).
• Status: Finished. Lack comments.
• Functionality (refer to individual functions for details)
  • Calculate RSA with partial/non-partial correlation. Can also define timewindow and step increment for RSA.

Regression_Analysis.py

• Description Two classes for performing liner regression.
• Status: Finished.
• Functionality (refer to individual functions for details)
  • Recursively compute linear regression on all dimensions
  • A group regression wrapper for regressional ERP analysis

RSA_plot.py

• Description A group of functions that uses matplotlib for plotting
• Status: In Progress. Lack comments; need to rearrange some functions for simpler interactions.
• Functionality (refer to individual functions for details)
  • Customized functions for plotting 1D/2D correlation data
  • Customized functions for plotting correlation matrix
  • Customized functions for plotting dendrograms

PPData_alt.py

• Description A class for simple preprocessing of EEG and variable data. Suitable for importing several datasets in our lab.
• Status: Finished.
• Functionality (refer to individual functions for details)
  • Import data from txt/csv files
  • Clear pre-labeled artifact trials
  • Reassign bin labels to EEG data
  • Produce classical ERP

Clustering_Auxiliaries.py

• Description A group of functions for clustering/dimension analysis
• Status: Finished. Lack comments.
• Functionality (refer to individual functions for details)
  • Perform Principal Coordinate Analysis (PCoA, a.k.a. classical MDS) on a given representational dissimilarity matrix (RDM)
  • Analyze the principal coordinates by correlating results with variables
  • Recover full tree structure from the result of sklearn AgglomerativeClustering function. The tree is arranged as a binary tree, and each parent cluster is formed by merging the two children sub-clusters. A dictionary coding the clusters at each layer of the tree in the form of mask is also returned for easy evaluation (if at some level a cluster reaches its terminal stage and contains only 1 member, the cluster label of the cluster will be inhereited in the deeper layer).
  • Plot dendrograms (some obtained from scipy examples)
  • Evaluate performances of a group of hierarchical cluster

CPerm.py

• Description A class for performing 1D cluster permutation test
• Status: Finished.
• Functionality (refer to individual functions for details)
  • 1D cluster permutation

Auxiliaries.py

• Description A group of simple auxiliary functions
• Status: Finished.

Main functions:

Note that the toolboxes I presented here belong to an ongoing project, and some changes made in the toolboxes might not be reflected in the main functions. As a result, some of these main functions might call a non-existent function or an outdated version of a modified class function. I will make specific main function as examples for interacting with the toolboxes once the project is finished.

EmSingle_ERP.py

Main function for EmSingle Dataset. Include functions for preprocessing and producing ERP using PPData_alt, functions for producing regressional ERP using Regression_Analysis, functions for clustering permutation test using CPerm, and plotting functions using RSA_plot. Also include functions using spatio-temporal cluster permutation from the MNE toolbox.

RSA_clustering.py

Main function for performing clustering/dimension analysis for EmSingle dataset. Include functions for PCoA and associated analysis, functions for producing, plotting, and evaluating hierarchical clusters, and plotting function. Mostly rely on variables_processing_alt, Clustering_Auxiliaries, and RSA_plot.

RSA_driver.py

Main function for performing RSA for EmSingle dataset. Note that this is an old driver. Include functions for preprocessing using PPData_alt, functions for RSA using variables_processing_alt, functions for clustering permutation test using CPerm, and plotting functions using RSA_plot.

multi_cond_RSA.py

Main function for performing RSA for EmSingle dataset.Include functions for preprocessing using PPData_alt, functions (including interpolation attempts) for RSA using variables_processing_alt, functions for clustering permutation test using CPerm, and plotting functions using RSA_plot.

phil_semantics.py

Similar to multi_cond_RSA.py, except designed for Phil's kiloword dataset.