On Tue, 27 Mar 2001, Rohini wrote:
> Hi
>
> I intend to calculate correlations between equity pairs. This could number
> 20,000 to 30,000
> pairs that need to be examined. I would run 3 different caluclations to find
> their correlation over daily, weekly and monthly price data. So I'm looking
> at between 60,000 and 90,000 correlation calculations with each time series
> consisting of about 100 points. The speed that this can be done is of
> greatest importance and hopefully we would be able to run
> this every day or at least once or twice a week. I will be using a PIII
> 900 MHz computer with 256 MB RAM. Can anyone tell me how long this will take
> to run if I'm using S-Plus 2000?
I get 10,000 in 54 seconds on a 1GHz machine, not using much memory. In
any case, if speed is critical, one can write C code to do this part and
link it in to S-PLUS.
Of course you should do your own tests, as I am not entirely clear I did
what you might have expected, and accessing the series will take some time.
--
Brian D. Ripley, ripley@stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272860 (secr)
Oxford OX1 3TG, UK Fax: +44 1865 272595
|