Hi Daniel, In some cases (like yours), TProfile::GetBinError resets the bin error to the average bin error in the profile. This happens with some combinations of small statistics or ratio error/bin_content. I have disabled this default action in the new version in CVS and added a new static function TProfile::Approximation to enable it on demand. see: http://root.cern.ch/root/htmldoc/TProfile.html#TProfile:Approximate Rene Brun Daniel CUSSOL wrote: > > Hello rooters, > > I would like to show you a small problem when a TProfile is displayed. > In the joint ProfTest.root file, there is a TProfile named "prof" which > was > filled with very small Y values (typically around and below 1.e-4). > When I plot it without scaling, I get uniform error bars which have > no correlations with the real error bars. When I scale it with the > command > prof->Scale(100.) and draw it, the errors bars are correctly displayed > for the small X values, but incorrectly diplsyed for high X values. When > I apply an additional scaling via tp->Scale(10) (the original spectrum > is now scale by a factor 1000) and draw it, the error bars are correctly > displayed. It is as if a threshold on error values was set for the error > bars display. Is this "normal"? Could it be eventually corrected? > > Thanks for your help > > Cheers > -- > > Daniel CUSSOL > > LPC Caen > > Boulevard du Marechal Juin > > 14050 CAEN CEDEX > > e-mail : cussol@in2p3.fr > > Tel : +33-(0)2-31-45-29-73 > > FAX : +33-(0)2-31-45-25-49 > > -------------------------------------------------------------------------------- > Name: ProfTest.root > ProfTest.root Type: unspecified type (application/octet-stream) > Encoding: base64
This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:12 MET