iraf.net

>From n.r.todd@ic.ac.uk Tue Jan 28 10:47:41 1997
Date: Tue, 28 Jan 1997 17:43:14 +0000 (GMT)
From: Neal Todd <n.r.todd@ic.ac.uk>
To: Frank Valdes <valdes@noao.edu>
Subject: Thank you.Frank,
just got your second message as I was about to send the one below
(email tennis!). I agree with what you say and for my simulations I
should make sure my pixel counts are large. For such simulations of model
fitting where things like chi-squared are calculated I do think the ranbuf
default needs to be changed by the user.Hi Frank,
Thanks for your help. If you've got my second email by now
you'll know that I wasn't looking into the problem hard enough and that
it was really a problem with the randon number generator. For example, if
artdata.ranbuf=1000 (default) and you make up an image with background of
10000 and add poisson noise you get a pretty ugly looking histogram even
with a large number of pixels. If ranbuf=0 its okay. As the time saved is
quite minimal by setting a buffer a comment I'd make is that for the benefit
of thickies like me, in a future help page there's an explicit note that
default value of 1000 is going to give unrealistic noise (it certainly
messes up my chi-squareds!). Again, thanks very much for your speedy and comprehensive help.Regards, Neal.