T-test for two z-normalized matrices in MATLAB

Dear Experts,

I have two correlation matrices (X and X1) , each of it is 25x25 size . I need to z-normalized it and then compare them with t-test with a false discovery rate of 0.01, to see if the difference significant.

For z-normalization I used :

Z = zscore(X,1)

Z2 = zscore(X1,1)

May you help me , to find way to applied t-test with a false discovery rate of 0.01 to my new matrices Z and Z2.

Thanks so much in advance!

Answer 1

[Samayochita]

Hi IM,

I understand that you want to compare two 25×25 correlation matrices (X and X1) elementwise and find out if they differ significantly.

If you only have two 25x25 matrices (not multiple samples per entry), you cannot do a t-test.

Alternatively, you could use a “permutation test” (which allows element-wise testing), followed by FDR correction.

• Compute the difference between X and X1 (element-wise).

• Use a permutation test: Randomly shuffle the matrix labels many times and recompute the differences to build a null distribution.

• Compare the observed differences to the null distribution to get p-values.

• Apply False Discovery Rate (FDR) correction (e.g. Benjamini-Hochberg) at 0.01 to identify significant differences.

Hope this helps!