simulation1.m

% Simulations for normally distributed data
% It is expected that all methods perform well
% with Pearson's being the best

mkdir('normal_no_outlier')
cd('normal_no_outlier')

n = 20;
p= 6;
SIGMA = eye(p);
SIGMA(SIGMA==1) = abs(randn(p,1));

tic
[r_alpha,t_alpha,p_alpha,vr,vt,vp]  = MC_corrpval(n,p,'Skipped Pearson');
time = toc
save results

parfor MC = 1:1000
    fprintf('running simulations MC %g\n',MC)
    X = mvnrnd(zeros(1,p),SIGMA,n);
    [r(:,MC),t(:,MC),pval(:,MC)] = skipped_Pearson(X,[],'ECP',5/100);  
end

FWE_p = length(find(sum(pval<=rst_hd(vp,0.05),1)))/1000;
FWE_r = length(find(sum(r>rst_hd(vr,0.95),1)))/1000;
FWE_t = length(find(sum(t>=rst_hd(vt,0.95),1)))/1000;
save results

parfor MC = 1:600
    fprintf('running simulations MC %g\n',MC)
    X = mvnrnd(zeros(1,p),SIGMA,n);
    p_alpha2 = MC_corrpval(n,p,'Skipped Pearson',5/100,[],diag(cov(X)));
    [~,~,~,h2(MC,:)] = skipped_Pearson(X,[],'ECP',5/100,p_alpha2);
end
FWE2 = sum(h2)/600;
save results