Hi,
when I profile plots, I get sometimes zero errors for bins with just one
entry. However, documentation says, this shouldn't happen for neither of
options "","s","i". I looked at the source code of
"TProfile::GetBinError(Int_t bin)" function on the web and this case
seems not to be implemented there. Anyway, that causes the problem in the
fits, because the zero error forces the function to go through the point
and ignores the others. Would it be possible to have a Fit option to
ignore points with zero errors?
Miro
I'm working with version 2.25/03 on SuSe Linux
Macro demonstrating this feature:
{
TNtuple *ntuple = new TNtuple("ntuple","ntuple","x:y");
for (Int_t i=0;i<11;i++) {
ntuple->Fill(gRandom->Rndm(),2*gRandom->Rndm());
}
TH2D *histo = new TH2D("histo","my histo",20,0.,2.,20,0.,2.);
ntuple->Draw("x:y>>histo");
TProfile *histo_pfx = histo->ProfileX();
for (Int_t i=0;i<=histo_pfx->GetNbinsX()+2;i++) {
printf("bin=%d, entries=%d, content=%f,
error=%f\n",i,histo_pfx->GetBinEntries(i),histo_pfx->GetBinContent(i),histo_pfx->GetBinError(i));
}
histo_pfx->Draw();
}
Output: (part only)
bin=5, entries=0, content=0.000000, error=0.000000
bin=6, entries=0, content=0.000000, error=0.000000
bin=7, entries=1, content=0.850000, error=0.000000
bin=8, entries=1, content=0.650000, error=0.000000
bin=9, entries=1, content=0.050000, error=0.000000
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:34 MET