Fig. 1

Flow chart illustrating the steps of the spca_randtest. The first step on the top panel shows one permutation that is used to obtain one value of fi + and fi-. To assess the statistical significance of global either local patterns permutations are repeated x times to obtain empirical distributions of fi + and fi- that are thus compared to the observed values of fi + and fi-. If at least one of the two is significant, the second step of the test exploits the eigenvalue distribution recorded over the permutations to obtain an empirical p-value for each eigenvalue, starting from the most positive (or most negative). As the first eigenvalue is significant in comparison with a chosen threshold, the following is tested and compared to a more stringent threshold (Bonferroni correction) until a non-significant eigenvalue is found and the routine stops