z = atanh(r) ;

works also with n-dimensional arrays

The link provided by Habib Muhammad Nazir Ahmad may confuse: that is the Fisher procedure to compare correlation coefficients, but it is obsolete.

If you wish to compare correlation coefficients, use bootstrapped effect sizes (https://it.mathworks.com/matlabcentral/fileexchange/68238-bootes-compare-correlations?focused=67ae44d8-a24e-43b9-8fc5-11814f92e785&tab=function).

This formula may be useful:

t_r1 = 0.5*log((1+r1)/(1-r1));

t_r2 = 0.5*log((1+r2)/(1-r2));

z = (t_r1-t_r2)/sqrt(1/(n1-3)+1/(n2-3));

here r is the input variable and n is the number of samples.

https://in.mathworks.com/matlabcentral/fileexchange/44658-compare_correlation_coefficients

https://www.statisticshowto.datasciencecentral.com/fisher-z/

Bijay Kumar Guha , Simone Di Plinio , Habib M. N. Ahmad Thank you.