How can add the rows in n-by-m matrix that have the same indicator in the last column?

Hi,

How can add the rows in n-by-m matrix that have the same indicator in the last column?

For example: 10-by-3 matrix, the third column is the indecator:

253638  3309  2
223389  3309  2
255968  3309  2
243943  33309  2
245568  63309  1
242490  93309  3
268464  23309  2
272665  53309  1
241206  83309  1
259310  13309  2

I want to add all rows (all values in first column with the same indicator) with ( all values in second column with the same indicator).

Answer 1

[Jos (10584)]

X = [253638  3309  2
223389  3309  2
255968  3309  2
243943  33309  2
245568  63309  1
242490  93309  3
268464  23309  2
272665  53309  1
241206  83309  1
259310  13309  2]

OUT = cell2mat(arrayfun(@(k) [sum(X(X(:,3)==k,[1 2]),1) k], unique(X(:,3)),'un',0))

[Abdulatif Alabdulatif]

Thanks for a single line answer!!

Could you add another column which contains how many data points belong to each class?

[Jos (10584)]

Juts modify the function used by arrayfun

OUT = cell2mat(arrayfun(@(k) [sum(X(X(:,3)==k,[1 2]),1) k sum(X(:,3)==k)], unique(X(:,3)),'un',0))

Answer 2

[Azzi Abdelmalek]

You can use accumarray function

A=[253638  3309  3
223389  3309  2
255968  3309  2
243943  33309  2
245568  63309  1
242490  93309  3
268464  23309  2
272665  53309  1
241206  83309  1
259310  13309  2]
[ii,~,jj]=unique(A(:,3))
out=accumarray(jj,(1:numel(jj))',[],@(x) {sum(A(x,1:2),1)})
out=[ cell2mat(out) ii]

[Walter Roberson]

That would be tricky to do without one accumarray call per column.

[Abdulatif Alabdulatif]

The first command is work!

the next two command show the following error:

Index exceeds matrix dimensions.

Error in @(x){sum(A(x,1:2),1)}

Answer 3

[Walter Roberson]

unique() to find the unique indicators and the number of indicators. Allocate an zero array of appropriate size to hold the results. Loop over the unique indicators finding the rows that match and totaling them and storing that in the results array.

accumarray() is not suitable for your purposes because accumarray cannot operator on row vectors of values.

Alternately if you convert the data into a table, you could perhaps use varfun()

[Abdulatif Alabdulatif]

Could you provide a poosible loop commands for the example?

Thanks

[Walter Roberson]

uinds = unique(YourMatrix(:,end));
ninds = length(uinds);
results = zeros(ninds, size(YourMatrix,2)-1, class(YourMatrix));
for K = 1 : ninds
  ind = uinds(K);
  results(K, :) = sum( YourMatrix( YourMatrix(:,end) == ind, 1:end-1) );
end

[Abdulatif Alabdulatif]

I got the following error for the second line of the loop. The marix name is y

Index exceeds matrix dimensions.

Error in kmc (line 46)
results(K, :) = sum( y( y(:,end) == ind, 1:end-1) );

what is the issue in this case?

[Abdulatif Alabdulatif]

It was my machine problem and the code works prefectly.

Thanks