How can I sort a matrix according to column IDs and identify same combinations?

Hi,

I edit the question: I have matrix M with six columns and n rows:

M = [HomeID, WorkID, SchoolID, ShopID, timeS, timeP];
for example (lets say it is already sorted):
M = [101, 201, 301, 401, 5.46, 7.83;
    101, 201, 301, 401, 3.63, 4.52;
    101, 201, 301, 401, 2.785, 2.99;
    102, 205, 301, 401, 3.53, 8.5;
    102, 205, 301, 402, 3.48, 8.9];

I want to make it in the form of:

 M2 = [101, 201, 301, 401, 3.958, 5.113;
       102, 205, 301, 401, 3.53, 8.5;
       102, 205, 301, 402, 3.48, 8.9];

I want to make calculation for each unique pair of SchoolID-WorkID combination within each unique pair of HomeID-WorkID combination and produce a matrix which will have no multiple Home-Work-School-Shop registers but the unique ones with the average values instead [M2(1,4)=(M(1,4)+M(2,4)+M(3,4))/3].

Thanks,

Iro

[Matt Kindig]

It is a little confusing how your data is structured. Please provide some sample data, as well as your target output for your sample data. Without seeing your data, it will be a bit difficult to recommend a general solution.

[Iro]

Thank you Matt, I edited the question, hope it makes more sense now.

[José-Luis]

Why is [102,205] not averaged since there are two values for it?

[Iro]

Because the School-Shop combination which comes next is different.

[Zhang lu]

102, 201, 301, 401, 3.53, 8.5; ？ what is it from?

[Iro]

Hi Zhang lu..Does it really matter? It is just random numbers I used to illustrate my question. :)

Answer 1

[Jonathan Sullivan]

Accumarray would come in handy here. It's a rather useful function, but might take a little bit of reading to understand it completely. Basically it takes in a list of coordinates (subscripts) and the values associated with them. It then does an operation on all those values that have the same subscripts, and outputs the resulting value as an entry in a matrix. In this case, you want to take the mean of those values that have the same HomeID WorkID SchoolID and ShopID.

Here's how we do it:

% Get the unique combinations, and the indexes
[HomeSchoolWorkShop_ID,~,y_ind] = unique(M(:,1:4),'rows');
x_ind = 1:size(M,2)-4;

% Make a list of the indexes and values
[X_Ind,Y_Ind] = meshgrid(x_ind,y_ind);
Inds = [Y_Ind(:) X_Ind(:)];
vals = reshape(M(:,5:end),[],1);

% Take the average
M2 = [HomeSchoolWorkShop_ID accumarray(Inds,vals,[],@mean)]

[Iro]

Hi Jonathan, thanks for your answer. I did it and works! *Accumarray *looks good for my case if it can accommodate also more complex calculations, I will take a careful look. Cheers! /Iro