Collecting statistics for time dependent data?

[Luke Stras]

On Thu, Dec 09, 2004 at 12:53:49PM +0100, Raimondo Giammanco wrote:
> Hello,
> 
>  I was wondering if there is a way to compute "running" statistics with
> gsl.
> 
>  Consider a time dependent simulation where at each time-step you
> compute a vector/matrix of data, and you want to compute the 
> sample mean, but using only one additional vector for data storage.

To the best of my knowledge, GSL doesn't have any routines to this.
However, I would recommend:

@Article{west79,
  author =       {D.H.D. West},
  title =        {Updating Mean and Variance Estimates: An Improved 
                  Method},
  journal =      {Comm. ACM},
  year =         1979,
  volume =       22,
  number =       9,
  pages =        {532-535},
  month =        {September}
}

@Article{chan79,
  author =       {T.F. Chan and J.G. Lewis},
  title =        {Computing Standard Deviations: Accuracy},
  journal =      {Comm. ACM},
  year =         1979,
  volume =       22,
  number =       9,
  pages =        {526-531},
  month =        {September}
}

I used to have some (non-GSL) code lying around to implement West's
method, but I can't find it anywhere.  It wasn't very complex, that's
for sure.

-- 
Luke Stras <stras@utias.toronto.edu>
"The meek can have the Earth; the rest of us have other plans" 
  --Henry Spencer