This folder contains figure source data for the figures presented in the Results section of:

RAA Ince, BL Giordano, C Kayser, GA Rousselet, J Gross and PG Schyns
“A statistical framework based on a novel mutual information estimator utilizing a Gaussian copula”

Fig11A.mat

• ks1000 : KS test ground-truth (1000 trials) [time x sensors]
• Icop100 : Gaussian-copula MI (100 trials) [time x sensors]
• tt100 : t-test (100 trials) [time x sensors]
• gtsig, Icop100sig, tt100sig : p=0.01 statistical significance for the above measures from permutation test + max statistics
• time : time index
• chanlocs : EEGLAB channel structure

Fig11C.mat

This includes confusion matrices from multiple repitions of calculations of a number of statistics:

• Ib2 : binned MI, 2 bins
• Ib4 : binned MI, 4 bins
• Ib8 : binned MI, 8 binss
• Icop : gauss-copula MI
• ks : Kolnogorov-Smirnov test
• t : t-test

The results variables are:

• trials_cm : structure with an element for each of the statistics. Each element is a 4d array [2 x 2 x repetitions x trials used], the confusion matrix (compared to ground truth) for each repitition and data size
• trials : number of trials used (last axis above)
• corrupt_cm : structure with an element for each of the statistics. Each element is a 4d array [2 x 2 x repetitions x data corruption level], the confusion matrix (compared to groun truth) for each repitition and level of corruption
• corrupt_prct : percentage of corrupted trials (last axis above)

Fig12A.mat

• Icop2d : MI in 2d planar gradient [sensors x lags]
• Icopamp : MI in planar gradient amplitude (pythagorean sum) [sensors x lags]
• Icopdir : MI in planar gradient direction [sensors x lags]
• Icop2dsig, Icopampsig, Icopdirsig : p=0.01 statistical significance for the above measures from permutation test + max statistics
• delays : lags axis (ms)
• chanlocs : EEGLAB channel structure

Fig12C.mat

This includes confusion matrices from multiple repitions of calculations of a number of statistics:

• Ib2 : binned MI, 2 bins
• Ib4 : binned MI, 4 bins
• Ib8 : binned MI, 8 binss
• Icop : gauss-copula MI
• sp : Spearman’s rank correlation
• pe : Pearson correlation

The results variables are:

• length_cm : structure with an element for each of the statistics. Each element is a 4d array [2 x 2 x repetitions x data used], the confusion matrix (compared to ground truth) for each repitition and data size
• lengths : length of data used (last axis above)
• corrupt_cm : structure with an element for each of the statistics. Each element is a 4d array [2 x 2 x repetitions x data corruption level], the confusion matrix (compared to groun truth) for each repitition and level of corruption
• corrupt_prct : percentage of corrupted trials (last axis above)

Fig13A-E.mat

• time : time axis (ms)
• conderp : stimulus conditional ERPs [time x deciles]
• Ip, Iv, Ipv : Gaussian-copula MI in respectively raw signal (p for position), gradient (v for velocity) and the bivariate response considering raw signal and gradient jointly (pv)
• xIp, xIpv : cross-temporal interaction information for the raw signal (p) and the bivariate raw + gradient signal (pv)
• inttime : time index for interaction matrices (ms)

Fig13F.mat

• Iemerge : time course of novel MI emergence
• time : time axis (ms)

Fig13G.mat

• Icopamp : GCMI in planar gradient amplitude
• Icop2d : GCMI in 2d planar gradient
• Icop4d : GCMI in 2d planar gradient + temporal derivative of each component
• lages : stimulus-MEG lags (ms)