Thanks to a question from Bing Jiang (in China ?) I've noticed that in CIAO 3.1 (July 2004) the default behaviour of dmextract was changed so that STAT_ERR is set to 1+sqrt(N+0.75), Neil Gehrels' approximation for low count data. I disagree with this strongly since it overestimates the errors for reasonably bright sources. The old behaviour can be recovered by psetting error=gauss. The POISSERR keyword is also set incorrectly - being set to F for sqrt(N) errors and true otherwise.
UPDATE: Note that this doesn't matter for xspec normally because if POISSERR=T then sqrt(N) errors are assumed however running the spectrum through grppha will change the keyword to F so xspec uses the errors in the STAT_ERR column.
Keywords: CIAO
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment