Product of probabilities by group

Hi,

I have matrix that contains in the first column a group identifier and in the second column a probability. I want to generate, for each group identifier, the product of the probabilities that correspond to that group.

For example

X=[1 0.5;1 0.3; 1 0.4; 2 0.7; 2 0.4];

What I want is to generate a matrix

Y=[1 (0.5*0.3*0.4);2 (0.7*0.4)];

and I can't do it "by hand" because the number of groups and the number of probabilities is huge.

Any ideas? Thanks.

Answer 1

[Matt Fig]

Here is another method. This might work best if you have groups in increments of one because it uses the group number as the index of the return array.

Y = accumarray(X(:,1),X(:,2),[],@prod);

Y(I) is the product of the probabilities belonging to the Ith group. To understand this, look at:

X = [1    0.5000
     1    0.3000
     1    0.4000
     2    0.7000
     4    0.5000
     4    0.6000
     6    0.7000
     6    0.2000];  % No group 3 or 5.
Y = accumarray(X(:,1),X(:,2),[],@prod);

Since there is no group 3 or 5 in X, Y(3) and Y(5) are zero. If there are missing groups in your array, you can still get to the desired form like:

Z = [find(Y),Y(Y~=0)];

Answer 2

[the cyclist]

Here's one way:

[uniqueX1,~,index] = unique(X(:,1));

numberUniqueX1 = numel(uniqueX1);

Y = nan(numberUniqueX1,1);

for nx = 1:numberUniqueX1
    Y(nx) = prod(X(index==nx,2));
end

Answer 3

[Fernando]

Thanks for your help!