Hi Marc-Andre,
I have fixed this problem appearing when subtracting identical histograms
in the CVS source. I have also modified the computaion of the number
of entries in this case.
Rene Brun
Marc-Andre Pleier wrote:
>
> Dear Rooters,
>
> is there anybody who can comment on the following effect?
>
> Just copy the following example into your ROOT-session:
>
> form1 = new TFormula("form1","abs(sin(x)/x)");
> sqroot = new TF1("sqroot","x*gaus(0) + [3]*form1",0,10);
> sqroot->SetParameters(10,4,1,20);
> h1f = new TH1F("h1f","Test random numbers",200,0,10);
> h1f->SetFillColor(45);
> h1f->FillRandom("sqroot",10000);
> h1f->Draw();
> h1f->Add(h1f, -1.);
> h1f->Draw();
>
> So, what's happening is that you subtract two identical histograms from
> each other and get - as expected - an empty histogram as result BUT
> the mean and rms are still the same of the original histogram being
> subtracted from itself...
>
> I think, what happens is the following:
> If you subtract two histograms, their values for sumw (stats[0]) and sumwx
> (stats[2]) are added for the resulting histogram instead of being
> subtracted.
> This will thus result in a wrong mean value.
>
> The way ROOT behaves right now, you can get two identical
> (non-empty!) histograms having different MEAN and RMS.
>
> I tried the same in PAW - there everything looks fine, and the
> above example would result in an empty histogram with mean = rms = 0.
>
> I tried this for ROOT-versions
> 3.01/06 29 August 2001 and
> 2.24/05 25 May 2000.
>
> If there's a good reason for this, I'd really like to know.
>
> Cheers,
> Marc-Andre
>
> P.S.: I've posted this already some time ago, but got no answer.
> I still think this is a severe problem...
>
> -----------------------------------------------------------------------------
> Marc-Andre Pleier Tel. : +0049 6221 516506
> MPI Kernphysik Fax : +0049 6221 516603
> Postfach 103980 Email: Marc-Andre.Pleier
> D-69029 Heidelberg @mpi-hd.mpg.de
> -----------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:00 MET