Re: How to extract sequenced data from an array?

[fork]

new_user <ghostility <at> gmail.com> writes:

> 
> > Yes, xcorr looks much more appropriate than what I said. It looks like
> > you only want a vector. You may want to run xcorr in autocorrelation
> > mode on each column
> 
> > c1=xcorr(X(:,1));
> > c2=xcorr(X(:,2));
> 
> > then maybe correlate the two results
> 
> > c=xcorr(c2,c2);
> 
> > then look for peaks in c. not sure if this is the correct approach.
> > However, xcorr on a bigx2 array may well result in you having to
> > terminate the process again.
> 
> I run c1=xcorr(X(:,1)); on data sample, but graph is symmetrical, suggesting
I'm not seeing native data
> positions to track down where sequenced data ends:

If I remember correctly (check me -- I wish I did more of this stuff), you can
ignore the left half of the graph, even throw it out. Then a spike at x=3000
means there is some periodicity at 3000.  (I think xcorr gives you correlations
in each direction, and you don't need half of those).

> As it can be noticed from screen-shot I run gnuplot with "every 10" as I get
out of memory otherwise

If your data is discontinuous integers, you might miss periodicity this way.  

Have you tried using the maximum lag parameter?

> I'm just afraid of large data sets :)

And you are in meteorology ??  well, you'll grow to love them.