Automatically generate new matrices or arrays from unique values in matrix

Hello,

I'm moderately experienced with programming, and I cannot believe how challenging I am finding this hurdle...

Suppose I have the following matrix:

a = [1,1;1,2.5;1,3;2,2;2,2;2,2;3,10;3,11;3,14;3,19;4,9]

and the first column represents a unique site or subject that I need to analyze separately. How can I make, in this case, 4 new matrices or arrays, within the workspace so that I have products that look like this:

site1 = [1,1;1,2.5;1,3]
site2 = [2,2;2,2;2,2]

My real data is more advanced than this, of course, but I cannot seem to figure out how to separate the data into separate sub-data sets based on a unique values. This seems to be a simple problem but I've tried looking at for-loops, the command 'unique', etc. and am getting absolutely nowhere.

Thanks ahead of time,

Ryan

Answer 1

[Sean de Wolski]

B = accumarray(a(:,1),a(:,2),[],@(x){x})

Using it directly like this, the index will be the leading number. Why you would need to replicate it, I don't know but it could be done easily. I.e:

B=
  [3x1 double]
  [3x1 double]
  [4x1 double]
  [         4]

Thus the first 3x1 could get three ones because it's the first index, etc. Let me know if you need more clarification.

EDIT (from the cyclist): Be sure to see Stephen Cobeldick's comment to this answer, which solves this entirely.

[Stephen23]

It is also easy to use accumarray to get the entire rows, which is also simpler than the accepted answer:

>> a = [1,1;1,2.5;1,3;2,2;2,2;2,2;3,10;3,11;3,14;3,19;4,9];
>> C = accumarray(a(:,1),(1:size(a,1))',[],@(r){a(r,:)});
>> C{:}
ans =
     1     1
     1     2.5
     1     3
ans =
     2     2
     2     2
     2     2
ans =
     3    10
     3    11
     3    14
     3    19
ans =
     4     9

[the cyclist]

I've unaccepted my answer and accepted this one, which is indeed syntactically simpler. However, if one is having a difficult time understanding the underlying "magic" of accumarray, I would encourage you to look at my alternative solution, which exposes the algorithm more clearly.

Answer 2

[the cyclist]

There are many ways to slice this particular cat. Here is one way I typically use. I've tried to use descriptive variable names, to help you understand the logic of what I am doing.

a = [1,1;1,2.5;1,3;2,2;2,2;2,2;3,10;3,11;3,14;3,19;4,9];

[uniqueFirstColumnValues,indexToUnique,indexFromUniqueBackToAll] = unique(a(:,1));

numberUniqueValues = numel(uniqueFirstColumnValues);

subsetData = cell(numberUniqueValues,1);

for nu = 1:numberUniqueValues
    indexToThisUniqueValue = (indexFromUniqueBackToAll==nu);
    subsetData{nu} = a(indexToThisUniqueValue,:);
end

Note that each subset of data is a differently sized matrix, so I chose to store the results in a cell array.

[Ryan Utz]

Thanks! Seems to work...

[Rachel Terry]

ahhh, just what I needed! Thanks from 5 years later!