[R] functions for high dimensional integral

[Prof Brian Ripley]

On Fri, 6 Jun 2008, ZT2008 wrote:

> I need to compute a high dimensional integral. Currently I'm using the
> function adapt in R package adapt. But this method is kind of slow to me.
> I'm wondering if there are other solutions. Thanks.

What does 'high' mean?  Numerical quadrature will be slow in more than a 
handful of dimensions.

What to recommend depends on what you know about the function -- Evans & 
Swartz (2000) 'Approximating Integrals via Monte Carlo and Deterministic 
Methods' is a good reference on integration for statisticians.

But accurate evaluation of an integral in more than 2 or 3 dimensions is 
potentially a very computer-intensive task -- people spend days of CPU 
time using e.g. MCMC to do just that.

> Zhongwen

-- 
Brian D. Ripley,                  ripley at 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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595