Accuracy and error analysis (was Re: [Maxima] primes)

--- Stavros Macrakis  wrote:

> The simplest technique for estimating a result distribution from
> input distributions is Monte Carlo. But it can be computationally
> expensive....

I'm inclined to believe that it is useful to try to
identify special cases, and fall back on a catch-all
like Monte Carlo for anything else. I believe Lisp and 
Maxima are well-suited to this approach since it is 
easy to inspect expressions at run time.

Special cases include not only exact results (e.g., product
of lognormal distributions is lognormal and so on) but also
approximations which are faster or more accurate than MC,
for example, computing an approximation to the distribution
of a sum via a Fourier transform of discretized distributions.

For what it's worth,
Robert Dodier


		
__________________________________ 
Yahoo! Mail Mobile 
Take Yahoo! Mail with you! Check email on your mobile phone. 
http://mobile.yahoo.com/learn/mail