Extracting unique values

Question

I have data in two columns that looks as follows:

A               B

1,265848208     3
-0,608043611    0
-0,285735893    0
0,006895134     7
0               7
-0,004526196    7
0,176326617     10
-0,159688071    2
0,22439945      2
-0,991045044    1
0,178022324     1
-0,270967397    4
0,285849994     4
1,881705539     23
1,057184204     10
NaN             10

For all unique values in B I want to extract the corresponding value in column A and move it to a new matrix. I'm looking to then compute the mean of all the corresponding values in A and use as a dependent variable (weighted by no of observations per value in B) in a regression with the common value of B being the independent variable to reduce noise. Any help would on how to do this in Matlab (except running the regression) would be great!

Thanks

Oscar

Answer 1

Here is an efficient solution:

X = [
    1.265848208     3
    -0.608043611    0
    -0.285735893    0
    0.006895134     7
    0               7
    -0.004526196    7
    0.176326617     10
    -0.159688071    2
    0.22439945      2
    -0.991045044    1
    0.178022324     1
    -0.270967397    4
    0.285849994     4
    1.881705539     23
    1.057184204     10
    NaN             10
];

%# unique values in B, and their indices
[valB,~,subs] = unique(X(:,2));

%# values of A for each unique number in B (cellarray)
valA = accumarray(subs, X(:,1), [], @(x) {x});

%# mean of each group
meanValA = cellfun(@nanmean, valA)

%# perform regression here...

The result:

%# B values, mean of corresponding values in A, number of A values
>> [valB meanValA cellfun(@numel,valA)]
ans =
            0     -0.44689            2
            1     -0.40651            2
            2     0.032356            2
            3       1.2658            1
            4    0.0074413            2
            7   0.00078965            3
           10      0.61676            3
           23       1.8817            1