CCHcoeff

[mvonpapen]

added option in elephant.spike_train_correlation.cch named cross_corr_coeff that allows to normalize the CCH output to obtain a cross-correlation coefficient function according to Eggermont (2010) Pair-correlation in the time and frequency domain, Ch.5 in: Analysis of parallel spike trains, eds: S. Gruen, S. Rotter, Equation (5.10).

added a unit test in test_cross_correlation_histogram of test_spike_train_correlation.py that compares the output of CChcoef to the one of elephant.spike_train_correlation.corrcoeff for several displacements tau.

Joint work with @rgutzen

def _cross_corr_coef(cch_result, binned_st1, binned_st2):
# Normalizes the CCH to obtain the cross-correlation
# coefficient function ranging from -1 to 1
N = max(binned_st1.num_bins, binned_st2.num_bins)
Nx = len(binned_st1.spike_indices[0])
Ny = len(binned_st2.spike_indices[0])
spmat = [binned_st1.to_sparse_array(), binned_st2.to_sparse_array()]
bin_counts_unique = []
for s in spmat:
bin_counts_unique.append(s.data)
ii = np.dot(bin_counts_unique[0], bin_counts_unique[0])
jj = np.dot(bin_counts_unique[1], bin_counts_unique[1])
rho_xy = (cch_result - NxNy/N) / np.sqrt( (ii-Nx**2./N)(jj-Ny**2./N) )
return rho_xy

[coveralls]

Coverage increased (+0.03%) to 90.922% when pulling 376dbf9 on INM-6:CCHcoeff into 9316970 on NeuralEnsemble:master.

[alperyeg]

Looks good to me. Please add your names to the authors.rst for the kudos :)

[mvonpapen]

ready, woohoo!

[coveralls]

Coverage increased (+0.03%) to 90.922% when pulling e37c994 on INM-6:CCHcoeff into 9316970 on NeuralEnsemble:master.

[coveralls]

Coverage increased (+0.03%) to 90.922% when pulling e37c994 on INM-6:CCHcoeff into 9316970 on NeuralEnsemble:master.