pyGC
Non-parametric implementarion of Granger Causality in python based on Dhamala's paper "Estimating Granger causality from Fourier and wavelet transforms of time series data".
This package contains the module pyGC contains the function wilson_factorization for factorizing spectral matrices based on the implementation by Nedungadi et. al., in "Analyzing multiple spike trains with nonparametric granger causality" (direct translation from their Matlab code), and the function granger_causality to compute the causal influences between two signal.
pySpec contains functions to compute power spectrum and wavelet spectrum.
With our code we reproduce Dhamala's example, to run the code first do:
- sh generate_enviroment.sh to create data and out folders
- run ipython gc_fft.py to generate data to reproduze Figure 1 from Dhamala's paper
- run ipython gen_wvl_mat.py to save the wavelet matrix for the data
- run sbatch run.sh to generate data with time-frequency Granger Causality.
- run ipython plot_gc.py to generate figures.