Re: random variate from power exponential distribution:performances

[Giulio Bottazzi <giulio dot bottazzi at libero dot it>]

On Wed, 29 Sep 2004 17:28:10 +0100
Brian Gough <bjg@network-theory.co.uk> wrote:

> The source code for exppow.c says 
> 
>     "Scale factor chosen by upper bound on ratio at b = infinity.
>      This could be improved by using a rational function
>      approximation to the bounding curve. "
> 
> Maybe that would help improve the rejection method?
> 

Hi,
prompted by the previous message by Brian I started to look closer to
the code for the rejection method in gsl_ran_exppow. I confronted
the code with the original paper by Tadikamalla, and I noticed
some differences between the two. Then, I rewrote the code following
step-by-step the suggestion by Tadikamalla (ref. in my previous posts)
and I obtained a DRAMATIC improvement in performance! The rewritten
rejection methods are, when available, noticeably faster than the gamma
method. My idea is to prepare a revised version of the file exppow.c and
send it to Brian. Before, however, I'd like to make some more tests and
check a couple of things. To this purpose I have two requests:

Does anybody have access to an electronic version of the paper

Johnson (1979) "Computer Generation of the Exponential Power
Distribution", Journal of Statistical Computation and
Simulation,9,239-240

and be so kind to send me a copy?

AND

does anybody who know about recent papers/proposals/code about the
generation of power exponential pseudo random variables be so kind
to mail me the references? (I found something, but I'm almost sure I've
also miss something!)

Many thanks in advance,

	G.

-- 
Giulio Bottazzi                 PGP Key ID:BAB0A33F 

CAFiM  Center for the Analysis of Financial Markets
LEM          Laboratory of Economics and Management
Sant'Anna School for Advanced Studies, Pisa, Italy
Phone: (+39)-050-883365       Fax: (+39)-050-883344
email: bottazzi@sssup.it    www.sssup.it/~bottazzi/

Attachment: pgp00000.pgp
Description: PGP signature