David,
	I must respectfully disagree.  I don't know about Boinc, but in S@H
classic, there is a power threshold below which Gaussians will not be
reported, no matter what the "fit" may be.  I believe that the Gaussians
must also be of the correct duration, to match the results of the Arecibo
antenna "sweeping" across a point radio source in the sky as Earth rotates.
	I have run dozens of WUs that SETIspy clearly showed with no Gaussians at
all, just before I sent them in for the record.  Spikes can certainly be
just noise, and the occurrence of spikes would be meaningless without
duplicating the results later; even then, it could be just a natural radio
source, such as a pulsar.  
-- 
(Replies *will* bounce, unless you delete
the letter A from my email address)
Cheers,
Red

David Woolley wrote:

In article <KiXILDDM8YPDFxMn@virgin.net>,
Bill Jillians <> wrote:

1. No Gaussians

You can fit a Gaussian to random noise, it just would be a very poor
fit. You would have to specify how poor for this post to make any

Not in S@H.  It tries to fit so many gaussians to each work unit that
approximately one of them will be good enough to report to Berkeley.
That's true of all the detection modes; they are all calibrated for
the order of one false positive per work unit. What will stop gaussians
is an unacceptable angle range, that stops them being tested for.

Also note that there is no absolute amplitude information in the work
unit so connecting the feed to any noise source will produce similar
statistics.