Re: deleting duplicities, diag, equidistant data

[Paul Kienzle]

octave-forge has interp1 for interpolating on a regular grid, but
it does not average the y values for repeated x.

If you only have pairs that are duplicated, and your x and y are
already sorted then:

  dups = find(x(1:end-1)==x(2:end));
  x(dups)=[];
  y(dups,:) = (y(dups,:)+y(dups+1,:))/2;
  y(dups+1,:) = [];

Below is a general solution for averaging y values for repeated x
not already ordered based on sort and sparse tricks.

- Paul

function [unique_x,unique_y] = avgdups(x,y)
  [unique_x,i,j]=unique(x(:));
  n=length(x); unique_n=length(unique_x);
  c = columns(y);
  if n != unique_n,
    z=sparse(j*ones(1,c),ones(n,1)*[1:c],y,unique_n,c,'sum');
    nz=sparse(j*ones(1,c),ones(n,1)*[1:c],1,unique_n,c,'sum');
    unique_y = full(z)./full(nz);
  else
    unique_y = y(i,:);
  endif
end

%!test
%! x=[3.2,1.5,2.6,3.2,2.6,7.8]';
%! y=[  3,  4,  2,  5,  1,  1;
%!     10, 12,  8, 12,  5,  9]';
%! [x,y]=avgdups(x,y);
%! assert(x,     [1.5,2.6,3.2,7.8]');
%! assert(y(:,1),[  4,1.5,  4,  1]');
%! assert(y(:,2),[ 12,6.5, 11,  9]');

%!test
%! x=[3.2,1.5,2.6,7.8]';
%! y=[  3,  4,  2,  1;
%!     10, 12,  8,  9]';
%! [x,y]=avgdups(x,y);
%! assert(x,     [1.5,2.6,3.2,7.8]');
%! assert(y(:,1),[  4,  2,  3,  1]');
%! assert(y(:,2),[ 12,  8, 10,  9]');

On Feb 28, 2005, at 3:03 AM, Jiri Pachman wrote:

> thanks to all for answers to my previous questions
>
> Hi everybody,
> I have this simple problem. During data manipulation I create two 
> (sometimes 3) columns of data. First contains x values the others are 
> y, y2 values. It sometimes happens that I get two values of y are for 
> the same value of x (around 5 couples in matrix of 200 lines). This 
> makes problems in further calculations. I tried gnuplot´s smoothing 
> (smooth spline) function and it works well, taking the duplicate 
> values and making y average for particular x. Unfortunately I do not 
> know, how to extract this data in numeric format from gnuplot. I was 
> looking for something similar in octave, but without luck. Is there 
> something like this in octave?
>
> For now I am using the following script that finds the duplicate 
> values and deletes them, but I would prefer to average them. I also 
> tried to work with diagonal using matlabs diag function, but it seems 
> to work quite differently (that is why I use rather complicated way to 
> change 1 on the diagonal to 0).
>
> x = G12;
> vel = rows(x);
> indexy = zeros(vel);
> for i = 1:vel
> ind = x(:,1) == x(i,1);
> indexy(:,i) = ind;
> diamat(i,i) = 1;
> endfor
>
> #% -------- indexes of duplicate values -------------
> err = (indexy - diamat);
> max(err)
>
> #%--------------------- adding err columns creating index  ---------
> ind = err(:,1);
> for k=1:vel-1
> ind = ind + err(:,k+1);
> G12_nove = x(~ind, :);
> endfor
>
> I am also interested, if there is some elegant way how to create 
> equidistant data from nonequidistant data. I use polynomial fit and 
> then I calculate the new values in linspace or logspace depending on 
> the need.
>
> thanks for help
>
> Jiri
>
>
> --
> ***************************************************
> Ing. Jiri Pachman
> Katedra teorie a technologie vybusnin
> Univerzita Pardubice
> Studentska 95
> 532 10 Pardubice
>
> tel.:  +420-46-603-8018
> fax.: +420-46-603-8024
> ***************************************************
>
>
>
> -------------------------------------------------------------
> Octave is freely available under the terms of the GNU GPL.
>
> Octave's home on the web:  http://www.octave.org
> How to fund new projects:  http://www.octave.org/funding.html
> Subscription information:  http://www.octave.org/archive.html
> -------------------------------------------------------------



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------