Divide a dataset into subsets

Hello everyone

I have a dataset named chan56 which is 1619936x1 and a vector called markers which is 1x202. The vector markers has the events of the dataset chan56 and i want to split chan56 into subets that contain -1200 indexes from each event until +300. I tried

Data = []
DataMatrix =[zeros(1,length(markers)*1500)];

for n =1 :length(chanPOz)
    for i =1:length(markers)
        if markers(i) == chanPOz(n)
            T = chanPOz(n);
            Data = chanPOz(n-1200:n+300);
            DataMatrix = [DataMatrix  DataMatrix+Data];
        end
    end
end

But i cant take the DataMatrix

[Akira Agata]

I'm confused a little bit...

Could you provide a simple example ??

[Jos (10584)]

I think you want to create a matrix with 202 rows (=number of markers). Each row contains the data from the chan56 around this marker.

What does an element in the vector markers represent? An index into chan56?

[Greg Athanasiadis]

The elements in markerks contain the timestamps so the first element contain the 27965th point in chanPOz. I saw that i ve made a serious mistake in my code. I changed it to

newData = [];
Data =[];
 for i = 1:length(chanPOz)
    for n =1 :length(markers)
        if i == markers(n)
         Data = chanPOz(i-1200:i+300); 
         newData =[newData Data];
        end
    end
end

I want to create a new vector newData with every 1500 points near the timestamp of vector markers. [-1200....marker(i)......+300]

[Jos (10584)]

I am a little confused by the if statement. Am I right to assume that each value of markers is a direct index into chanPOZ?

[Greg Athanasiadis]

Yes indeed. As an example Assume that i have my data

data = [ 1 5 6 12 15 18 16 12 10 17 25 31 35 36 35 33 32 24 29 35 40 36 33 25]

the vector markers has the indexes of events 18-36-40

markers = [6 14 21]

so i want a subset with [markers(i)-2 markers(i)+1]

In my original dataset i want [markers(i)-1200 markers(i)+300]

first dataset= [-1200th -1199th -1198th ......markers(1).....+299 +300th]

Second dataset= [-1200th -1199th -1198th ......markers(2).....+299 +300th]

...... Last dataset= [-1200th -1199th -1198th ......markers(n).....+299 +300th]

Answer 1

[Jos (10584)]

This is what you're after, I think. No need for a loop over each point in chanPOz

N = numel(markers) ;
% pre-allocate the result:
newData = zeros(N, 1501) ;  % 1 extra !! [x-1200:x+300] are 1501 values
% loop over each marker
for n = 1:N
   idx = markers(n) ; % the n-th index into chanPOz
   newData(n,:) = chanPOz(idx-1200:idx+300); % get the data and store this in the n-th row
end

[Greg Athanasiadis]

Thank you both for your answers!!!!

I ran Jos solution because i have Matlab 2015b but i had problems with the limits so i add an if statement so the loop started from 1200 and ended on N-300. Matlab ran the loop but the matrix new Data was still with zeros

Answer 2

[Guillaume]

Even a single loop is not needed. A double loop is a complete waste of time.

Assuming R2016b or later,

assert(all(markers > 1200 & markers < numel(chanPoz) - 300), 'some markers too close to the edges');
indices_to_extract = markers(:) + (-1200:300); %create a matrix of indices
extracteddata = chanPoz(indices_to_extract.');  %extract relevant data as a matrix

Each column of extracteddata is the -1200 to 300 points around the corresponding marker. If you do want it as a vector (but why?) then you can then do:

extracteddata = reshape(extracteddata, 1, []);

If you're on a version earlier than R2016b, then the indices_to_extract line becomes

indices_to_extract = bsxfun(@plus, markers(:), -1200:300);

[Jos (10584)]

+1 I still need to get used to the "new" expansion behaviour of + ...

[Greg Athanasiadis]

I got Error using + Matrix dimensions must agree.

[Guillaume]

I clearly wrote:

If you're on a version earlier than R2016b, then the indices_to_extract line becomes

indices_to_extract = bsxfun(@plus, markers(:), -1200:300);