How to use cumsum on a segment of a large matrix

I have a matrix

 pos_ht=[1 5 5; 1 7 16; 1 9.10 6; 1 12 8; 2 5.5 5; 2 7 5; 2 9.7 6; 3 12 8; 3 4.7 5; 3 7 5; 3 9.10 6];

which represents just an tiny example of the whole matrix, for explanation purposes. The first column represents the week number. I want to cumulate column 3 for each week segment and output all values in a single vector. So the resulting vector should look like this

L=[5; 21; 27; 35; 5; 10; 16; 8; 13; 18; 24]

I am so close to the solution, but have problems with combining, or rather depositing the answers in a single vector:

L=size(pos_ht); 
gr=L(1,1);
L=zeros(gr,1); %space allocation

for k=1:3
findrange=find(pos_ht(:,1)==k)
anfang=min(finderange) %anfang=beginning
ende=max(finderange)   %ende=end
L=cumsum(pos_ht(anfang:ende,3))
end

Right now the for-loop delivers

L =
        5.00
       21.00
       27.00
       35.00
L =

          5.00
         10.00
         16.00
L =

          8.00
         13.00
         18.00
         24.00

I have tried adding more "linkage" (basically adding index parentheses everywhere L(k) or findrange(k), but that got me nowhere...

I have a feeling this is really simple! :D Any help will be much appreciated

Answer 1

[Star Strider]

This seems to do what you want:

pos_ht=[1 5 5; 1 7 16; 1 9.10 6; 1 12 8; 2 5.5 5; 2 7 5; 2 9.7 6; 3 12 8; 3 4.7 5; 3 7 5; 3 9.10 6];

for k1 = 1:max(pos_ht(:,1))
    C{k1,:} = cumsum(pos_ht(k1 == pos_ht(:,1),3));
end
L = cell2mat(C);

[Inga Beyers]

Perfect - Thanks for the timely response! However, could you clarify some things for me, for future reference... So

cumsum(pos_ht(k1 == pos_ht(:,1),3))

is basically cumulating the values for each week as k1 progresses towards its maximum. No problem there...

However, how does the C{k1,:} fit them all together? And why is it suddenly a cell array?

Thanks a lot!

[Star Strider]

My pleasure!

It’s a cell array becasue it’s easier to store vectors of different lengths in a a cell array. The two-dimensional array reference forces ‘C{k1,:}’ to be a column vector of (Nx1) vectors (with varying N). This makes it easier to concatenate them using cell2mat, creating the vector you want.