Operating on One Column Based on Another

So I've run into a buzzsaw of a problem. I have a m x n numeric matrix (no strings, etc.) and two columns are of interest. Column 1 looks like this:

A = [1;1;1;1;2;2;2;3;3;3;3;3;4;4;4;5;5;5]

Next, column 2:

B = [1;2;3;4;5;6;7;8;9;10;11;12;13;14;15;16;17;18];

What want to do is look at where values of A are equal and operate on those values in the same position in B. In other words, A(1:4) = 1 so I want to add together (or standard deviation, variance, etc.) B(1:4). Then, I want to move on to the next set of equal values of A and operate on the corresponding positions of B, and so on. Once values have been added together, I want to compare the result against a tolerance value and if the sum falls outside of the value, I want to delete those rows.

The problem I'm running into is that all of the comparison methods I've tried automatically dismiss the last value in B. So, where A(5:1) doesn't equal A(4:1), I'll get a sum, let's say, for B(1:3) instead of B(1:4). For example, one method I tried was:

a = unique(matrix(:,col_A))
tol = 5;
for i = 1:length(matrix)
    idx = find(matrix(i,col_A) == a);
    if isequal(matrix(i),idx)
        std_dev = std(matrix(m,B));
        if std_dev < tol
            matrix(matrix(m,:)) = [];
        end
    end
end

At first, because of how isequal works, the algorithm discards A(4,1) doesn't equal A(5,1). I understand why this is happening, but I can't think of a viable solution. I tweaked it a bit, but now when I run the algorithm as written above, nothing happens at all.

If I was unclear at any point, don't hesitate to let me know. Any help you all can provide would be most appreciated!

Answer 1

[Stephen23]

You should look into accumarray:

>> A = [1;1;1;1;2;2;2;3;3;3;3;3;4;4;4;5;5;5];
>> B = [1;2;3;4;5;6;7;8;9;10;11;12;13;14;15;16;17;18];
>> S = accumarray(A,B,[],@sum)
S =
    10
    18
    50
    42
    51
>> S(A)
ans =
    10
    10
    10
    10
    18
    18
    18
    50
    50
    50
    50
    50
    42
    42
    42
    51
    51
    51

If the values of A are not consecutive integers starting from one, then use unique first and use its third output argument.

[Jeremy]

Stephen - First, thank you very much for your reply. This looks like exactly what I need. Here's a follow-up question if you have time: these are the vectors I'm actually working with

A = [10517;10517;10517;10517;29272;29272;29272;29272;29272;35942; ...
35942;35942;35942;35942;39017;39017;39017;39017]

B = [-25;37;-34;69;-20;-21;-22;-23;-25;5;6;8;5;4;-17;-20;-25;-23]

In order to make things work, I had to rewrite A to be:

new_A = [1;1;1;1;2;2;2;2;2;3;3;3;3;3;4;4;4;4]

Leaving it in the original configuration produced a 39017x1 matrix of 0s, which was extremely unhelpful. Further, I had to hardcode new_A so things would work as advertised. Is there a clever way I can substitute values of new_A into A so I don't have to hardcode anything? Also, I wasn't following your advice regarding the use of the 'unique' function, so if that has an impact on things, I wouldn't know at this point.

Thanks a lot for your time and help!

[Stephen23]

[~,~,newA] = unique(A,'stable')

[Jeremy]

You are my new hero. That is perfect and I sincerely appreciate your time. Take care and have a good weekend!