Compute correlations in 3D arrays

5 views (last 30 days)
%Random matrices
A=randi(100,374,374);
A_ = eye(size(A,[1 2]));
A(ones(size(A))&A_)=NaN;
B=randi(100,374,374);
B_ = eye(size(B,[1 2]));
B(ones(size(B))&B_)=NaN;
The following code computes correlation coeficient and p value from matrices A, B:
nn=374;
temp= ~eye (nn);
ii_all_conn = find(temp>0);
ii_uptri_conn = find(triu(temp,1)> 0);
ii_lotri_conn = find(tril(temp,-1)> 0);
%Corr plots up entries
figure, plot(A(ii_uptri_conn), B(ii_uptri_conn),'o');
[r,p]= corr(A(ii_uptri_conn), B(ii_uptri_conn));
title(['Upper connections - r = ' num2str(r) ' (p ' num2str(p) ')']);
%Corr plots low entries
figure, plot(A(ii_lotri_conn), B(ii_lotri_conn),'o');
[r,p]= corr(A(ii_lotri_conn), B(ii_lotri_conn));
title(['Lower connections - r = ' num2str(r) ' (p ' num2str(p) ')']);
Can I compute the same correlation and p-value in multidimensional arrays? E.g.
A_3D=randi(100,374,374,10);
B_3D=randi(100,374,374,10);
In the output, the first r and p values would correpond to the Pearson coeficient of A(:,:,1), B(:,:,1). and the tenth r and p values correpond to the Pearson coeficient of A(:,:,10), B(:,:,10)

Accepted Answer

Dyuman Joshi
Dyuman Joshi on 7 Nov 2023
Run a for loop through the 3rd dimension -
A_3D = randi(100,374,374,10);
B_3D = randi(100,374,374,10);
s = size(A_3D,3);
[ru, pu, rl, pl] = deal(zeros(s,1));
for k = 1:s
[ru(k), pu(k), rl(k), pl(k)] = correlation(A_3D(:,:,k), B_3D(:,:,k));
end
%Upper triangle values
[ru pu]
ans = 10×2
0.0029 0.4429 0.0032 0.3950 0.0069 0.0684 -0.0013 0.7218 -0.0056 0.1426 -0.0027 0.4775 -0.0026 0.4976 -0.0055 0.1459 -0.0012 0.7442 0.0031 0.4171
%Lower triangle values
[rl pl]
ans = 10×2
0.0007 0.8508 -0.0015 0.6969 -0.0113 0.0028 0.0065 0.0878 0.0001 0.9798 0.0064 0.0892 -0.0038 0.3129 -0.0029 0.4408 -0.0012 0.7495 0.0043 0.2520
function [Ru, Pu, Rl, Pl] = correlation(A, B)
A = modify(A);
B = modify(B);
temp= ~eye(size(A,[1 2]));
%% Logical indexing is faster than find()
ii_uptri_conn = triu(temp,1)> 0;
ii_lotri_conn = tril(temp,-1)> 0;
[Ru,Pu] = corr(A(ii_uptri_conn), B(ii_uptri_conn));
[Rl,Pl] = corr(A(ii_lotri_conn), B(ii_lotri_conn));
end
function in = modify(in)
temp = eye(size(in,[1 2]));
in(ones(size(in))&temp) = NaN;
end

More Answers (0)

Categories

Find more on Creating and Concatenating Matrices in Help Center and File Exchange

Products


Release

R2023b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!