grouping data based on the times a value is repeated

I have data in array form with big number of rows and 12 columns. One of the column has an increasing array but not in a stable way (e.g. 1 1 1 1 1 2 2 2 2 3 3 4 4 4 4 4 4 4 4 5 6 6 6 etc)I want to organize data so I have all rows with 1 repetition of the value, 2 repetitions and so forth without disrupting the order of the sequence between each number. So I want to have an array that in the beginning was like this:

[1  45  67]
[1  23  89] 
[1  90  110]
[2  41  52]
[2  51  76]
[2  77  88]
[2  13  14]
[3  545 111]
[3  242 53]
[3  80  23]
[4  11  22]
[4  14  26]
[4  16  77]
[4  11  943]

and I want it to be like that:

[1  45  67]
[1  23  89] 
[1  90  110]
[3  545 111]
[3  242 53]
[3  80  23]

and

[2  41  52]
[2  51  76]
[2  77  88]
[2  13  14]
[4  11  22]
[4  14  26]
[4  16  77]
[4  11  943].

The data is all numerical and all rows have values in all columns.

Is there a way to do this with MATLAB?

Thank you very much.

Answer 1

[Guillaume]

m = [1  45  67; 1  23  89; 1  90  110; 2  41  52; 2  51  76; 2  77  88; 2  13  14; 3  545 111; 3  242 53; 3  80  23; 4  11  22; 4  14  26; 4  16  77; 4  11  943]

numreps = diff(find(diff([0;m(:, 1);0])));  %or use any histogram function
[~, neworder] = sort(numreps);
splitm = mat2cell(m, numreps, size(m, 2));
reorderedm = cell2mat(splitm(neworder))

Other ways of obtaining numreps:

numreps = accumarray(m(:, 1), 1);  %only if m(:, 1) is integer from 1 to x with no gap.
numreps = histcounts(m(:, 1), 'BinMethod', 'integers'); %only if m(:, 1) is integer from 1 to x with no gap.
numreps = histcounts(m(:, 1), [unique(m(:, 1)); Inf]);

[Georgios Tertikas]

thank you very much. can you tell me how i can find the values that the number of rows change? (eg 3 rows, then 4rows and so forth)

[Guillaume]

If I understand correctly what you're asking, replace the 2nd line with

[sortedreps, neworder] = sort(numreps);

to find the length of each run

runlength = diff(find(diff([0; sortedreps; 0])))

[Georgios Tertikas]

thank you very much

Answer 2

[Image Analyst]

Have you tried taking the histogram of the first column?

[Georgios Tertikas]

can you be more specific please?

[Image Analyst]

I know you already have a working example so this is probably too late, but here's the code:

% Define sample data.
m = [...
1  45  67
1  23  89
1  90  110
2  41  52
2  51  76
2  77  88
2  13  14
3  545 111
3  242 53
3  80  23
14  11  22  % Note: increases but there is a gap from 3 to 14.
14  14  26
14  16  77
14  11  943]
% Get histogram of first column.
counts = histcounts(m(:,1))
% Sort them and keep track of order
[sortedCounts, sortOrder] = sort(counts, 'ascend')
output = []; % Initialize output to null.
% Put in rows into output in the desired order.
for k = 1 : length(sortOrder)
  % Skip any with a count of 0 because those numbers are missing.
  if sortedCounts(k) == 0
    continue;
  end
  % Find out which rows have a count of "k".
  rowsToAppend = m(:, 1) == sortOrder(k);
  % Now append only those rows (with a count of k) onto output.
  output = [output; m(rowsToAppend, :)];
end
output % Print to command window to see results.

Note that this code is robust enough to still give the output even if there is a gap in the increasing numbers, like going from 3 to 14 instead of from 3 to 4. It still works.

Not as compact as Guillaume's but perhaps it's more understandable about how it works. It's well commented.

[Georgios Tertikas]

i tried it but the histogram function doesnt work on my data for some reason

[Image Analyst]

You must have a really old version of MATLAB. You can use hist() or histc() instead.

[Guillaume]

In my answer, under "Other ways of obtaining numreps", I showed two different ways of using histcounts. Because of the automatic binning you can't just pass the column of number.

Answer 3

[Georgios Tertikas]

i meant it works but it doesnt show the results i need