function [r p LOOpredictor LOOmaps] = LOO(zmaps,measure,covariates)
%Leave one out analysis using an M x N matrix of zmaps and an N x 1 matrix
%of measurements to predict.
s = size(measure,1);
h = waitbar(0,’wait’);
if nargin==2
for i=1:s
waitbar(i/s);
tmp = zmaps;
tmp2 = measure;
tmp(:,i) = [];
tmp2(i) = [];
LOOmaps(:,i) = corr(tmp’,tmp2,’rows’,’complete’);
end
for i=1:s
tmp = zmaps(:,i);
LOOmap = LOOmaps(:,i);
% LOOmap(LOOmap>0)=0;
tmp2 = corr(tmp,LOOmap,’rows’,’complete’);
% tmp2 = corr(tmp,LOOmaps(:,i),’rows’,’complete’);
LOOpredictor(i) = atanh(tmp2);
end
[r p] = corr(LOOpredictor’,measure,’rows’,’complete’);
else
for i=1:s
waitbar(i/s);
tmp = zmaps;
tmp2 = measure;
tmp3 = covariates;
tmp(:,i) = [];
tmp2(i) = [];
tmp3(i,:)=[];
LOOmap = partialcorr(tmp’,tmp2,tmp3,’rows’,’complete’);
tmpmap = zmaps(:,i);
LOOpredictor(i) = atanh(corr(tmpmap,LOOmap,’rows’,’complete’));
end
[r p] = partialcorr(LOOpredictor’,measure,covariates);
end