Review materials for Tressoldi et al 2014, versions 1&2
Supplementary materials for my review of Tressoldi et al. 2014. "Brain-to-Brain (mind-to-mind) interaction at distance: a confirmatory study" version 1.
Arbitrarily chosen stimulus labels can be decoded equally well as the real stimulus:
Here I analysed raw data from the "Receiver" in Pair 1 using non-linear SVM classification with a RBF kernel (k-nearest neighbour classifier with k=3 produces similar results). The black boxcar line indicates the stimulus labels used to decode (1=signal, 0=silence). The red dots denote classification for each time point. The overall decoding accuracy is shown above each plot.
Several sets of stimulus labels were used:
ActualStimulus: the actual stimulus labels used in the experiment with one signal period in the middle of the session.
FakeStimulus1: an arbitrarily chosen signal period that does not overlap the true signal period.
FakeStimulus2: arbitrarily chosen stimulus labels that alternates between signal and silence over the course of the session.
FakeStimulus3: arbitrarily chosen stimulus labels with the first half of the session defined as stimulus and the second half as silence.
FakeStimulus4: arbitrarily chosen stimulus labels in which the session was divided into 2s periods each of which had a 50% probability of being defined as stimulus.
FakeStimulus5: randomly chosen stimulus labels in which each sample was randomized with 50% probability to be assigned as stimulus. For display purposes only a short section of the whole decoded time series is shown but the accuracy is based on the entire time series. Unlike for the previous examples of arbitrary stimulus labels, for high temporal frequency labels like this decoding does not work.