Ranking row values of matrix without adjustment

Hello, I have a matrix that is 8000x600 and I need an output matrix of the same dimensions, containing the ranks of each element per row. I have used tiedrank but this averages the two ranks of equal observations and assigns to both that rank, which is not useful for me.

To make it simple, if I have [ 7 6 2 9 2] I would like the output to have [4 3 1 5 2] so that the value "2" gets 2 different ranks and not the average as [4 3 1.5 5 1.5].

Thank you in advance for your help

Answer 1

[Jos (10584)]

for a vector:

V = [ 7 6 2 9 2]
[~,r] = sort(V) 
rankV(r) = 1:numel(V)

for a matrix, you can loop over the columns

M = magic(5)
[~,r] = sort(M)
szM = size(M) 
rankM = zeros(szM) % pre-allocation
ridx = 1:szM(1) 
for k=1:szM(2)
   rankM(r(:,k),k) = ridx 
end

Other approaches using, for instance, ndgrid and sub2ind are also fine but less readable, I think.

[David Goodmanson]

For a matrix you can also do

[~,a] = sort(M);
[~,rankM] = sort(a);

[Jos (10584)]

Yes, sure, I just overlooked that. Thanks, David!