Comments on: People

By: nestor

nestor — Wed, 16 Dec 2009 11:13:10 +0000

Thanks for the comment Vinay. Yes I see that. I am just intrigued at the
way people bin and report their findings. A 0.2 sigma shift for a
couple of sources would create a nearly normal
distribution. This experiment has only detected the Crab nebula at 6 sigma
and now it’s making a significant leap in its detection limits albeit at a
purported 2 sigma. The equivalent in X-rays would be an experiment that only
detects is the Crab at 5 sigma, sees nothing in between and then
publishes a paper claiming two sigma detections for the faintest sources
in the Chandra deep field under the same argument. Would you believe
such a claim?. It looks statistically sound but there is counterintuitive here
that just doesn’t compute. Thanks again.

By: vlk

vlk — Fri, 04 Dec 2009 17:30:36 +0000

It seems that what they are doing is to compute a significance for every source in their catalog and compare the distribution of significances to what would be expected from a Gaussian distribution. They find 7 sources at significances >2sigma. Seeing 7 out of 27 (25%) at >2sigma is not very probable, so, in conjunction with the coincidence of these sources with known pulsars, they claim the detection. 2 sigma implies that 5% of the time a random deviation can produce numbers greater than that threshold. Because it is a two-sided distribution, on the upper side that is a 2.5% probability, which, for 27 sources, implies 0.6 sources are expected to exceed that number.

I think the methodology is fine. It is a combination of marginal detections superposed on known catalog sources that makes the cut.

By: nestor

nestor — Fri, 04 Dec 2009 09:03:56 +0000

has anyone read this paper?. it has been accepted by ApJ Letters.

http://arxiv.org/abs/0912.0386

can multiple 2-sigma detections a detection make without stacking?. i would appreciate any comments. a baffled scientist.

Thanks,
nestor