Suppose I have the following array:
A=[1 2
1 4
2 6
3 5
5 7
1 9
2 7];
What is the most efficient (preferably one-liner) way in MATLAB to automatically obtain a matrix which holds the average value of the second column elements for every unique element in the first column? The result for A above would be:
B=[1 mean([2 4 9])
2 mean([6 7])
3 mean([5])
5 mean([7])];
Thanks in advance!
unique can generate the first column while the powerful accumarray equipped with the mean function will calculate the second (nonzeros is called for non-existent indices like 4 in this case):
>> B = [unique(A(:,1)),nonzeros(accumarray(A(:,1),A(:,2),[],@mean))]
B =
1.0000 5.0000
2.0000 6.5000
3.0000 5.0000
5.0000 7.0000
As noted in the comments by dan-man, the above approach only works if the first column always has integer values (I guess I implied this use-case from the given example; apologies). You can use unique to get around the shortcoming as follows:
[Aunq,~,Aind] = unique(A(:,1));
B = [Aunq,accumarray(Aind,A(:,2),[],@mean))];
Further pointed out was the fact that due to how accumarray works internally, calling mean directly is not as efficient as performing the calculation more directly:
B = [Aunq,accumarray(Aind,A(:,2),[],@sum)./accumarray(Aind,1,[],@sum)]
