Software for Directed Brain Network Studies of Intracranial EEG Data

The software for implementing the Bayesian hierarchical model proposed in the paper:

  • Li, H, Wang, Y, Yan G, Sun, Y, Tanabe, S, Liu, C, Quigg, M, and Zhang, T (2021). A Bayesian State-Space Stochastic Block Model for Mapping Epileptic Brain Networks. Journal of the American Statistical Association, 116(536):1637-1647.

    Matlab Codes for Modular Oscillatory Dynamic Directional Model

    The following codes implement the Bayesian inference of the modular oscillatory dynamic directional model (MODDM) proposed in the paper:

  • Zhang, T, Sun, Y, Li, H, Yan, G, Tanabe, S, Miao, R, Wang, Y, Caffo, B, and Quigg, M (2020). Bayesian Inference of a Directional Brain Network Model for Intracranial EEG Data. Computational Statistics and Data Analysis, 144:106847.

  • BMODDM.rar: The compressed folder for BMODDM codes.



    • Matlab Codes for Bayesian Modular and Indicator-based Dynamic Directional Model (BMIDDM)

    The following codes implement the Bayesian modular and indicator-based dynamic directional model proposed in the paper:

  • Zhang, T, Yin, Q, Caffo, B, Sun, Y, and Boatman-Reich, D (2017). Bayesian Inference of High-Dimensional, Cluster-Structured Ordinary Differential Equation Models with Applications to Brain Connectivity Studies. The Annals of Applied Statistics, 11(2): 868-897.

  • BMIDDMfuns.zip: The compressed folder for BMIDDM codes.



    • Matlab Codes for Scalar-on-Image Regression Using Ising-DP Prior

    The following matlab code implements Bayesian scalar-on-image regression using Ising-DP prior for image data, proposed in the paper:

  • Li, F*,Zhang, T*, Wang, Q, Gonzalez, MZ, Maresh, EL, and Coan, JA (2015). Spatial Bayesian variable selection and grouping in high-dimensional scalar-on-image regressions. The Annals of Applied Statistics, 9:687-713.
    • *Equally contributing authors.

  • DataGeneration.r: The r code that generates data defined on 10x10x10 lattice.
  • InitSimuGamma.m: The matlab function that assigns initial values to indicator variables \gamma.
  • InitSimuWH.m: The matlab function that assigns initial values to the weights W_h of values generated from DP.
  • neighbor_list.m: The matlab function that creates the list of neighbors for each voxel given voxels' spatial locations.
  • Vicinity.m: The matlab function that creates the vicinity matrix used in the Ising prior given voxels' spatial locations.
  • CondPostGamma.m: The matlab function that simulates selection indicator variables \gamma from its posterior conditional distribution.
  • SimuPostLambda.m: The matlab function that simulates coefficients of covariates from their posterior conditional distribution. In the paper, the covariate is just the intercept of scalar-on-image regression.
  • SimuPostSigma2.m: The matlab function that simulates the variance of regression errors from its posterior conditional distribution.
  • SimuPostThetaH.m: The matlab function that simulates the DP values \theta_h from its posterior conditional distribution.
  • SimuPostWH.m: The matlab function that simulates the weights W_h of the DP values from their posterior conditional distribution.
  • SimuPostZh.m: The matlab function that simulates the DP cluster indicators Z_h of predictros from their posterior conditional distribution.
  • master.m: The matlab code that conducts the posterior simulations given data.

    • Note 1: The file master.m would call all the defined functions for posterior simulations, and output three files: Simu_gamma.txt file contiaining simulated \gamma, Simu_ThetaH.txt file containing simulated DP values, and Simu_Zh.txt containing simulated DP cluster indicators of predictors.

    Note 2: To run master.m, the user needs to provide three files: X.txt (an nxp predictor matrix), Y.txt (the response vector of length n), and Location.txt (a 3xp matrix denoting the spatial locations of p predictors).

    Note 3: The user may run r file DataGeneration.r to generate data. This file will output three files: Location.txt of 1000 predictors defined on 10x10x10 lattice, Y.txt file of response variable of length 104, and X.txt file of 104x1000 pedictor matrix.

    Note 4: The vaues of hyperparameters a & b for Ising prior are set for predictors located on 3d lattice. The user may use different sets of hyperparameters depending on data properties.