Calculating Kendall's tau

Hi, I'm trying to calculate Kendall's tau using the 'corrcoef' command. My code so far is:

A_col = A_short'

idx = find(~isnan(B)+~isnan(A_col)==2);

[RHO,PVAL] = corrcoef(B(idx),A_col(idx),'type',Kendall)

The last line produces the error 'Undefined function or variable 'Kendall'. According to matlab help this is a valid value so how can I specify that I want a Kendall tau correaltion as oposed to a Pearson correlation?

Please help. Thanks

Answer 1

[Wayne King]

Hi Anna, 'Kendall' is not an option of corrcoef(). It is an option for the function corr(), which is part of the Statistics Toolbox.

So, this will work:

    x = randn(30,4);
    y = randn(30,4);
    y(:,4) = sum(x,2);
    [r,p] = corr(x,y,'type','Kendall');

provided you have the Statistics Toolbox

[Anna]

Hi Wayne, thank you that works brilliantly! I was worried that using corr instead of corrcoef would give me a huge matrix of r and p values but there's only one thankfully! Cheers :)

Answer 2

[VENKATA PHANIKRISHNA B]

a = [15, 18, 21, 24, 27]' ;

b = [25, 25, 27, 31, 32]' ;

[RHO,PVAL] = corr(a,b,'Type','Pearson');

fprintf('Pearsons correlation: %.3f \n',RHO);

fprintf('Pearsons P: %.3f \n',PVAL);

[RHO,PVAL] = corr(a,b,'Type','Spearman');

fprintf('Spearman correlation: %.3f \n',RHO);

fprintf('Spearman P: %.3f \n',PVAL);

[RHO,PVAL] = corr(a,b,'Type','Kendall');

fprintf('Kendall correlation: %.3f \n',RHO);

fprintf('Kendall P: %.3f \n',PVAL);

Pearsons correlation: 0.953

Pearsons P: 0.012

Spearman correlation: 0.975

Spearman P: 0.033

Kendall correlation: 0.949

Kendall P: 0.033