Getting stats results with a script

This is useful to script when many files/results need to be read at certain locations. This is illustrated here on an ANCOVA on ERP (channel*time): we want to know the effect of covariates where and when the condition effect is significant.

For Spatial-temporal clustering

1 - Call limo_stat_values for the results

Here using spatial-temporal clustering with a cluster defining threshold at p 0.05

LIMO = load(fullfile(pwd,'LIMO.mat')); LIMO = LIMO.LIMO;
[pvalues, mask] = limo_stat_values('Condition_effect_1.mat', 0.05, 2, LIMO);

The resulting mask here has 2 clusters and it is labelled in from 0 to N. NOTE if using max or tfce, there is no cluster and mask is a binary image (0 and 1) only

Ncluster = length(unique(mask))-1; % -1 because 0 does not count
figure; imagesc(mask); title('significant cells'); 

2 - Get summary statistics from clusters

cond_clusters = limo_get_summary('Condition_effect_1.mat',mask);
cov1_clusters = limo_get_summary('Covariate_effect_1.mat',mask);
cov2_clusters = limo_get_summary('Covariate_effect_2.mat',mask);

3 - Report, making use of LIMO.mat to get some information.

for c=1:Ncluster
    roi       = mask == c;        % get the cluster c

    % cluster info
    sig_times = find(sum(roi,1)); % find non 0 in time (sum over space)
    sig_chan  = find(sum(roi,2)); % find non 0 in space (sum over time)
    fprintf('cluster %g: starts at %g, stops at %g\n',c,...
        LIMO.data.timevect(min(sig_times)),LIMO.data.timevect(max(sig_times)))
    channames = arrayfun(@(x) x.labels, LIMO.data.chanlocs(sig_chan), 'UniformOutput', false);
    for n=1:length(channames)
        if n==1
             fprintf('cluster %g: covers channels %s ',c, channames{n})
        elseif n == length(channames) 
            fprintf('%s\n',channames{n})           
        else
             fprintf('%s ',channames{n})
        end
    end    
    fprintf('mean F values condition:%g cov1:%g cov2:%g\n', ...
        cond_clusters(c).mean,cov1_clusters(c).mean,cov2_clusters(c).mean)
end

What if we use TFCE - ie no clusters

1 - Call limo_stat_values for the results

[pvalues, mask] = limo_stat_values('Condition_effect_1.mat', 0.05, 3, LIMO);

2 - Get summary statistics

Here, we are 'making up' clusters, looking at which cells are contiguous in terms of significance, this is not the same as spatial-temporal clustering but an appreciations of regional effects, which makes sense because tfce scores indicate that they belong to clusters. See how to report results here.

[cond_clusters,madeupclusters] = limo_get_summary('Condition_effect_1.mat',mask);
cov1_clusters = limo_get_summary('Covariate_effect_1.mat',madeupclusters);
cov2_clusters = limo_get_summary('Covariate_effect_2.mat',madeupclusters);

3 - Report

for c=1:(length(unique(madeupclusters))-1)
    roi       = madeupclusters == c;        % get the cluster c

    % cluster info
    sig_times = find(sum(madeupclusters,1)); % find non 0 in time (sum over space)
    sig_chan  = find(sum(madeupclusters,2)); % find non 0 in space (sum over time)
    fprintf('tfce regional effect %g: starts at %g, stops at %g\n',c,...
        LIMO.data.timevect(min(sig_times)),LIMO.data.timevect(max(sig_times)))
    channames = arrayfun(@(x) x.labels, LIMO.data.chanlocs(sig_chan), 'UniformOutput', false);
    for n=1:length(channames)
        if n==1
             fprintf('regional effect %g: covers channels %s ',c, channames{n})
        elseif n == length(channames) 
            fprintf('%s\n',channames{n})           
        else
             fprintf('%s ',channames{n})
        end
    end    
    fprintf('mean F values condition:%g cov1:%g cov2:%g\n', ...
        cond_clusters(c).mean,cov1_clusters(c).mean,cov2_clusters(c).mean)
end