Monday, December 20, 2004

Sporadic problem of large error bars in faked spectra

David Buote has had occasional problems when the first fakeit in a session produces a spectrum with some large error bars and hence small chi-squared. I was unable to reproduce the problem locally however inspection of a file that David produced shows that the STAT_ERR column is large for channels that are in groups but are not the start of the group. These channels should have zero counts and errors. Only the former is true. Inspection of the wrtfak.f code shows that the error array written is not initialized to zero unlike the count array. I've passed instructions back to David on how to fix this bug and now await the results of his tests.

UPDATE: David reports that the bug is fixed so placed the correct version of wrtfak on the
xspec bugs page as 11.3.1r.

Keywords: xspec

No comments: