5. Neural‐Behavioral Modeling - jasonnan2/Automated-Analysis-of-EEG GitHub Wiki

This page explains how to model relationships between neural features and behavioral variables using the NeurBehMdl framework. It applies to both ScalpObject and SourceObject. Sample script can be found at the bottom of both runERP.m or runSource.m

Example

% Simulate a behavioral table
behTbl = table();
rng(1234)
behTbl.Subject = {sublist1, sublist2};
behTbl.behVar1 = rand(1,length(behTbl.Subject)';

% Obj here can either be a scalpObject or sourceObject as long as the main analyze function has been run. 
% Extract significant neural data into a tabular format
obj = obj.calSigTbl('sig');
% Calculate the neural behavioral models
obj =obj.NeurBehMdl(neuralVar,behTbl,keyColumnName,baseModel,'modelName');
% Extracts p values
obj.neuralBehMdl.modelName_NeuralVarName = extractp(obj.neuralBehMdl.modelName_NeuralVarName,"NeuralVarName",1);

Output from scalpObject.neuralBehMdl.modelName_NeuralVarName

Model Formula Model Type NeuralVarNameneuralP
NeuralVarName_time1_alpha_Pz LinearModel (1×1) 0.49731
NeuralVarName_time1_theta_FC4 LinearModel (1×1) 0.36107
NeuralVarName_time1_theta_O2 LinearModel (1×1) 0.22291

📚 Functions

obj.calSigTbl(...)

Extracts significantly different neural measures and saves them into a table for each variable and time window. Can be done for both scalp or source objects. Must be run after obj.analyzeRoi() or obj.analyzeScalp()

Parameter Description
chanTypes (optional) A string specifying which channels/ROIs to include in the table:
• 'sig' (default) extracts only significant channels/ROIs
• 'all' includes all channels/ROIs

Output: Populates obj.results.sigValues.(variable).(time) with the resulting tables.

NeurBehMdl(neuralVar, behTbl, keyColumnName, baseModel, modelname)

Fits behavioral linear models that relate neural data to behavioral variables. Full model will be baseModel+neuralVar. e.g. if baseModel = behVar1 ~ 1 + and neuralVar = {'neuralVar1}, then the full model will be behVar1 ~ 1 +neuralVar1.

Parameter Description
neuralVar Cell array of neural variable(s) names to test. Must match variables defined in obj.info.variables.
behTbl A table containing behavioral data. Must include subject IDs in a column with a name that matches keyColumnName.
keyColumnName String specifying the column name in behTbl for subject IDs.
baseModel A base model string for fitlm (e.g., "behVar1 ~ 1 +"). The neural variable is automatically appended to this model specification.
modelname (Optional) A label to identify the model. If not provided, a valid field name is created by concatenating baseModel and the neural variable name.

Output: Saves the fitted models in obj.neuralBehMdl.(modelname_variable) for each neural variable and time window.

extractp(behModels, template, showall)

Extracts p-values corresponding to the neural predictors from the fitted behavioral models.

Parameter Description
behModels The table or struct output returned from NeurBehMdl containing the fitted models.
template A string used to identify the neural predictor in the model's coefficient names (e.g., 'NeuralVarName').
showall (optional) A flag (1 or 0). If set to 1 (default), all p-values are returned; if set to 0, only models with p < 0.05 are retained.

Output: Adds a new field/column named <template>neuralP to the behModels structure containing the extracted p-values.

⚠️ **GitHub.com Fallback** ⚠️