Hi Thomas, Well, obviously, you should not fit with a "gaussian" what is not a gaussian. You should define a function with the right variable. What I could do automatically is the creation of the histogram with the variable size bins computed with the logarithm. Rene Brun Thomas Bretz wrote: > > Hi Rene, > > I'm using a variable bin size histogram like you showed in your example. > So I get the graphical output, as I like it. But if I try to fit this > histogram with a gaussian the 'raw' data is fitted. But I want to fit > the logarithmic data (3, 4, 5 not 1e3, 1e4, 1e5) The gaussian I get > after calling TH1::Fit doesn't even look like a gaussian (for sure) when > log scale is enabled. > > I would like to have both advantages. The advatage of a nice display of > the x-axis and the advantage, that I can fit the logarithmic data > instead of the raw data. > > Regards, > Thomas. > > Rene Brun wrote: > > > I have a histogram which is filled with some data. The get a more > > > conviniet view of the x-axis I set the bins of the histogram so, that if > > > I call gPad->SetLogx() I get equidistant bins. Now I want to fit this > > > view with a gaussian, but TH1::Fit fits the original data instead of the > > > logarithmic data. > > > > > > Is it possible to get either the nice view of a log-scaled X-Axis if i > > > would use equidistant bins and fill the histogram with log(data) or is > > > it possible to tell the fitter, that it should fit the logarithmic data > > > instead of the 'real' data in the histogram?
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:05 MET