Why the p value is so small in mann-whitney test?

Hi,

I have two matrix with diferent sizes

A(9*16)

B (3*16)

I want to check if the data from A are signifcantly diferent from the data in B using MWT (since not normal distribution)

This is my procedure:

C=[A; B];

group= [ones(size(A)); 2 * ones(size(B))];

g=group(:);

D=C (:);

[p, h, stats] = ranksum (D,g)

is there something wrong? my p value is returned very small!

p =

9.3520e-121

h =

logical

1

stats =

struct with fields:

zval: -23.3666

ranksum: 62128

Roxan

Answer 1

[Jeff Miller]

I think you use ranksum like this:

[p, h, stats] = ranksum (A,B)

With your command, it is testing whether the A & B scores are different from 1 & 2 (the group codes)

[Roxan]

Oh, Ok, the group was used to deal with the matrix situation!

but in this case, perhaps I should first make the A and B vecor by

A=A(:)

B=B(:) and then use

[p, h, stats] = ranksum (A,B)

right?

Thanks

[Jeff Miller]

Yes, I think that should work.