Andy, Thanks for your reply. I see the point. Rene Brun Andy Haas wrote: > > Hi Rene, > I would in principle be happy to implement this. Getting the errors > correct is very tricky, however. Assumptions have to be made about > the distributions from which the graphs are derived, and any > correlations between them. For many problems, the errors can not be > calculated correctly without knowing the actual histograms from which > the graphs were created. An example would be the ratio of two > efficiency curves, which share a common histogram as a denominator. I > do not suggest implementing a 'deafult' error calculation which > assumes Poisson-distributed uncorrelated data, since many users would > not realize that this formula is incorrect for their particular > situation. > So, I think Tancredi really should use a loop, as it's not much code > anyways. > Cheers, Andy. > > --- Rene Brun <Rene.Brun@cern.ch> wrote: > > Hi Tancredi, > > > > We do not have a TGraphErrors::Divide function. However, after some > > gymnastic, > > you should be able to do the job via TGraphAsymmErrors::BayesDivide > > or the corresponding TGraphAsymmErrors constructor. > > See the description of this function at: > > > http://root.cern.ch/root/htmldoc/TGraphAsymmErrors.html#TGraphAsymmErrors:BayesDivide > > > > You will have to transform your TGraphErrors to histograms before > > using this > > function. > > A direct interface TGraphAsymmErrors::Divide taking two TGraphs as > > input > > could be implemented. > > I am ccying this mail to Andy Haas the author of BayesDivide. > > Andy, would you like to implement TGraphAsymmErrors::Divide ? > > > > Rene Brun > > > > Tancredi Carli wrote: > > > > > > Hello, > > > > > > I have two objects of type TGraphErrors, let's say TG1 and TG2 > > > and I would like to divide them, e.g. TG3=TG1/TG2 and get > > > the correct error treatment. Is there an function for this > > > operation which I overlooked ? > > > > > > ..or do I have to get the values at each point by GetPoint > > > make the division, calculate the error, create new objects > > > and set the points ? > > > > > > Thank you for you help. > > > > > > Tancredi Carli > > > > > > -- > > > > > > > > > +-------------------------------------------------------------------+ > > > | Tancredi Carli tancredi.carli@cern.ch > > | > > > | CERN, Division EP phone ++41 22 767 9606 > > | > > > | CH-1211 Geneva 23 Room 40-1-C06 > > | > > > > > > +-------------------------------------------------------------------+ > >
This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:09 MET