chop array and average

Hi!

I have a dataset 1000x10 that the nth colum is an index, i.e. either 0, 1 or 2 but in a non-regular sequence or length. The rest of columns are my data. A simplified example would be:

1 0

2 0

4 1

6 1

7 1

4 2

3 2

1 0

7 0

2 1

5 0

6 0

2 2

Where column 1 is the data and column 2 is the index column.

I would like to average the "blocks" of data where 0-index is continuous and ommit the data with 1 and 2 indexing. Therefore I want to average the data in the first 2 cells, the 8th with the 9th and the 11th with the 12th cell and obtain:

1.5

4

5.5

How to do this in matlab?

Many thanks in advance!

Answer 1

[Image Analyst]

You can simply call regionprops() if you have the Image Processing Toolbox.

data = [1    0
	2    0
	4    1
	6    1
	7    1
	4    2
	3    2
	1    0
	7    0
	2    1
	5    0
	6    0
	2    2] ;

% Find out where last column is 0.
mask = data(:, end) == 0;
% Measure mean values there in each region of 0s independently:
props = regionprops(mask, data(:, 1), 'MeanIntensity');
theMeans = [props.MeanIntensity] % Extract from structure array into vector.

If you want to do it over all columns, put into a loop

[~, numRegions] = bwlabel(mask);
numColumns = size(data, 2) - 1 % All columns except the last one.
theMeans = zeros(numColumns, numRegions);
for col = 1 : numColumns
    % Measure mean values there
    props = regionprops(mask, data(:, col), 'MeanIntensity');
    % Store all the means for this column into one row of theMeans.
    theMeans(col, :) = [props.MeanIntensity]
end
    

[Efhtymios Polatidis]

Thanks! Very compact and it works

Answer 2

[KSSV]

data = [1    0
2    0
4    1
6    1
7    1
4    2
3    2
1    0
7    0
2    1
5    0
6    0
2    2] ;

A = zeros(1,size(data,1)) ; 
A(data(:,2)==0) = data(data(:,2)==0,1) ; 


ii = zeros(size(A));
jj = A > 0;
ii(strfind([0,jj(:)'],[0 1])) = 1;
idx = cumsum(ii).*jj;
out = accumarray( idx(jj)',A(jj)',[],@(x){x'});
iwant = cellfun(@mean,out)

[Efhtymios Polatidis]

Thanks it works!