Calculate differences through an array, by means of local absolute references inside of a matrix. And saving results.

Dear Friends,

I would need to do the same as I asked in the question: "how to calculate differences inside an array , looping through a matrix and saving results" I do not want to repeat the question so as not to be repetitive.

I got this answer.

    [a,b]=unique(x(:,1),'stable');
    b=[b;size(x,1)];
    out=nan(size(x));
    for k=1:numel(b)-1
      i1=b(k)+1;
      i2=b(k+1)-1;
      out(i1:i2,:)=bsxfun(@minus,x(i1:i2,:),x(i1-1,:));
    end
    out(isnan(out(:,1)),:)=[];
    out=out(:,2:end)

Taking into account the results provided in this wire, below are shown only a sample of them see original file in:

https://www.dropbox.com/s/kpifg9z43x4me39/out.txt

       imput :

      306  305,952  304,683  302,265  300,219 (1)
      406  406,016  404,943  402,467  400,224 (2)
      506  506,038  505,182  502,518  500,878 (3)
      606  605,983  605,304  602,749  601,037 (4)
      706  705,992  705,459  703,378  701,678 (5)
      806  806,027  805,31  803,212  801,897 (6)
      906  906,015  905,542  903,513  902,139 (7)->jump
      306  306,008  304,977  302,45  299,846 (8)<-
      406  406,009  404,961  402,483  400,361 (9)
      506  506,003  505,208  502,728  500,684 (10)
      606  605,999  605,447  603,032  601,458 (11)
      706  706,012  705,41  703,335  702,125 (12)
      806  806,02  805,462  803,338  801,901 (13)
      906  906,014  905,25  903,571  901,725 (14)

    desired output:
    this should be the first group:

    100  100,064  100,26  100,202  100,005 (row2-row1)
    200 200,086  200,499  200,253  200,659 (row3-row1)
    300 300,031  300,621  300,484  300,818 (row4-row1)
    400 400,04  400,776  401,113  401,459 (row5-row1)
    500 500,075  500,627  500,947  501,678 (row6-row1)
    600 600,063  600,859  601,248  601,92  (row7-row1)->jump
    100  100,001  99,984  100,033  100,515 (row9-row8)-<
    200  199,995  200,231  200,278  200,838 (row10-row8)
    300  299,991  300,47  300,582  301,612 (row11-row8)
    400  400,004  400,433  400,885  402,279 (row12-row8)
    500  500,012  500,485  500,888  502,055 (row13-row8)
    600  600,006  600,273  601,121  601,879 (row14-row8)

the calculations, start substracting from the second row the first one. then, the from the third to the first one. as you can see in the output. It is like absolute references in Excel, but, changing this reference to the smallest value of the first columm. This first columnn is not used in the calculations but I think is useful to index the arrays. having the start in the smallest value, and ending the loop in the biggest value found in the columm/array.It would be good if it can be applied for more columms and more rows. I do not know if It is possible. Any help is welcome. I think the code to write , is quite similar as one is shown here.

Thank you very much for your help. and I hope your reply.

Answer 1

[Andrei Bobrov]

m = dlmread('path_your_out.txt');
t=[true;diff(m(:,1)) < 0];
out = m-n(cumsum(t),:);
out = out(any(out,2),:);

or

m = dlmread('path_your_out.txt');
s = size(m,2);
n = find(diff(m(:,1))<0,1,'first');
m1 = reshape(m,n,[],s);
m2 = bsxfun(@minus,m1(2:end,:,:),m1(1,:,:));
out = reshape(m2,[],s);