5.1 Simulated data

In this section we compare the results from PARAFAC and tensor-PICA on the different artificial data sets (A)-(E). For each method, we show unthresholded spatial maps, together with both estimated time-courses and the 'true' (data sets (A) and (B)) or best rank-1 approximation to the set of true (data sets (C) - (E)) time-courses. The accuracy in the subject domain is shown via bar plots with both the estimated (left bars) and true (right bars) weights for the three subjects. In each plot, we show the results of the estimated source with highest spatial correlation with the true sources. The boxplots signify how strong the correlation is relative to the correlation between all other processes and the true spatial maps. Multiple large correlation coefficients signify substantial cross-talk between estimated spatial maps.