Suggestion for the function bst_permtest

gianmarcoduma · February 24, 2021, 12:26pm

Hi everyone, I'm writing for a suggestion about the use of the function bst_permtest for the statistics. I have let's say 20 subjects with 100 channels and two conditions A and B. I only have one point per channel, therefore I have a 100x1 matrix for each subject per condition. I want to use this function since my data are not derived from a brainstorm analysis but from an inlab script, but I need to use permutation statistics.
In the bst_permtest is written

INPUTS:
% - A, B : [Ma x Mb...] Variables to test
% - TestType : {'ttest_equal', 'ttest_unequal', 'ttest_paired', 'wilcoxon_paired', 'wilcoxon', 'signtest'}
% ttest_equal : Student's t-value, equal variance
% ttest_unequal : Student's t-value, unequal variance
% ttest_paired : Student's t-value, paired data
% signtest : Sign of the differences (paired)

however I am not sure how to structure the input data. I tried to compose a separate matrix for each condition A and B. The matrix was structured channels x subjects therefore 100 x 20. But I received the error message:
Error using permute
ORDER must be a numeric array.

Error in bst_permtest (line 57)
A = permute(A, [dimPerm, dimData]);

How should I structrure the input data in this function?
Thanks

Francois · February 24, 2021, 5:34pm

The 100x20 data matrix is fine, but you probably have to change the other variables.

The easiest is probably to put a breakpoint at the beginning of bst_permtest and then run some simple test from the Brainstorm interface. It is called from here:

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/process/functions/process_test_permutation2.m#L208


        OPTIONS.TestType = 'absmean_unconstr';

    else

        bst_report('Error', sProcess, [], ['Cannot run this test on unconstrained sources:' 10 'you must compute the norm of the three orientations first.']);

        return;

    end

end





% === COMPUTE TEST ===

% Run the permutation test

[pmap, tmap, nA, nB] = bst_permtest(sDataA.(matName), sDataB.(matName), OPTIONS.TestType, OPTIONS.Dim, Randomizations, OPTIONS.TestTail, OPTIONS.isZeroBad);

% Finished processing

bst_progress('text', 'Saving the results...');



% Bad channels: For recordings, keep only the channels that are good in BOTH A and B sets

switch lower(sInputsA(1).FileType)

    case 'data'

        ChannelFlag = sDataA.ChannelFlag;

        ChannelFlag(sDataB.ChannelFlag == -1) = -1;

    case {'results', 'timefreq', 'matrix'}

        ChannelFlag = [];

gianmarcoduma · February 24, 2021, 6:58pm

Very helpful as usual. Now it works fine, thank you so much!!!