Find the indicies of the top 250 values of a Matrix

I performed a spearman correlation analysis that produced a Rho variable. The size of the Rho variable is:

sizeRho = size(Rho) = [983, 27471]

Now, I am trying to find the the indicies for the top 250 values of Rho.

Here is some of what I have tried:

sizeRho = size (Rho);
top = maxk(Rho,250);
[B,I] = maxk(Rho,250, 'ComparisonMethod', 'abs');
[top, ixt, jxt] = prctile(Rho, [90])
top = prctile(Rho, [99])
largest_vals = maxk(Rho(:), 250);
[largest_vals, indicies_t] = maxk(Rho(:), 250);

Any help and/or advice would be greatly appreciated. Thank you.

[Matt J]

Now, I am trying to find the the indicies for the top 250 values of Rho.

It looks like you already did so in this line,

top = maxk(Rho,250);

[Emma Davis]

The variable 'top' has 250 rows, but 27471 columns. If I want the indicies of the top 250 values contained in my matrix, 'Rho', shouldn't I have 250 rows and 250 columns?

I really appreciate your help!

size(top)

ans =

250 27471

[Torsten]

Why ? maxk returns a matrix whose columns contain the k largest elements of each column of A.

Thus the matrix has dimension 250x27471.

[Emma Davis]

Yes. I believe I was using the wrong function by using maxk, which is why I also tried using the prctile function. I did not want the 250 largest elements of each column, but rather the 250 largest elements of the entire matrix. I hope this is clearer now.

Answer 1

[the cyclist]

[Sorry if you saw the incorrect solution I posted and then deleted.]

If you want the max values and indices for the matrix overall (rather than each column), you need this syntax:

[B,I] = maxk(Rho(:),250, 'ComparisonMethod', 'abs');

Note the use of Rho(:) as an input, rather than just Rho.

The output I has the linear indices into M.

For example

M = magic(4)

M = 4×4

16 2 3 13 5 11 10 8 9 7 6 12 4 14 15 1

[B,I] = maxk(M(:),3, 'ComparisonMethod', 'abs')

B = 3×1

16 15 14

I = 3×1

1 12 8

[Emma Davis]

Thank you so very much! This worked!

Answer 2

[Chunru]

Not very clear on what you are asking. Here is a guess.

Rho = rand(5, 6)                                    % small matrix for illustration

Rho = 5×6

0.6787 0.1810 0.3615 0.8210 0.5301 0.1768 0.6672 0.1126 0.0277 0.1434 0.8380 0.4849 0.4192 0.4177 0.3058 0.5544 0.7236 0.3155 0.7835 0.1233 0.4060 0.2886 0.9057 0.2207 0.4596 0.8407 0.9688 0.6253 0.7939 0.7684

[B, I] = maxk(Rho(:), 7, 'ComparisonMethod', 'abs')    % top 7 of all Rho (not columnwise)

B = 7×1

0.9688 0.9057 0.8407 0.8380 0.8210 0.7939 0.7835

I = 7×1

15 24 10 22 16 25 4

[row, col] = ind2sub(size(Rho), I);
[row col]

ans = 7×2

5 3 4 5 5 2 2 5 1 4 5 5 4 1

[Emma Davis]

This is perfect! Thank you so much! I apologize that my question was not clearer.