Averaging replicate data in Matlab, multiple variables

Question

I'm trying to average replicate data in MATLAB and running into some difficulty. The variables are depth, Var1, Var2. Sometimes there is a replicate in Var1, sometimes there is a replicate in Var2, sometimes there is a replicate for both Var1 and Var2 for a given depth. So the matrix might look something like this:

1   0.2 1,
2   0.5 3,
2   0.7 NaN,
3   0.1 5,
3   0.7 6,
4   0.3 4,
...

depth is the unique identifier so I would like to create a matrix with [depth, Var1, Var2] that looks like this:

1   0.2 1,
2   0.6 3,
3   0.4 5.5,
4   0.3 4,
...

The function accumarray would work if I had an n-by-2 matrix, but this is n-by-3. Any recommendations on how to proceed?

Answer 1

This should work

a=[1 0.2 1; 2 0.5 3; 2 0.7 NaN; 3 0.1 5; 3 0.7 6; 4 0.3 4];
depths = unique(a(:,1));
b=nan(length(depths),3);
for ct = 1:length(depths)
    b(ct,:)=mean(a(a(:,1)==depths(ct),:),1,'omitnan');
end

result

b =

    1.0000    0.2000    1.0000
    2.0000    0.6000    3.0000
    3.0000    0.4000    5.5000
    4.0000    0.3000    4.0000

Answer 2

A bit naive implementation with accumarray which loops over the variables.

A = [1   0.2 1
2   0.5 3
2   0.7 NaN
3   0.1 5
3   0.7 6
4   0.3 4];

result = zeros([numel(unique(A(:,1))) size(A,2)]);
result(:,1) = unique(A(:,1));
for ii = 2:size(A,2)
    result(:,ii) = accumarray(A(:,1),A(:,ii),[],@mean);
end