Hi Richard,
You should specify the option "UO" in your case.
Your histograms have about 8 per cent of entries overflowing.
Root try to detect weighted histograms by comparing the number of
entries with the sum of bin contents.
Rene Brun
On Fri, 21
Mar 2003,
Dubois, Richard wrote:
> One more try. I think my mail got deleted due to ~150 kB of attachments! So, see them at
>
>
>
> http://www-user.slac.stanford.edu/richard/DropBox/KS/ <http://www-user.slac.stanford.edu/richard/DropBox/KS/>
>
>
>
> Richard
>
>
>
> -----Original Message-----
> From: Dubois, Richard
> Sent: Friday, March 21, 2003 2:25 PM
> To: 'Roottalk@pcroot.cern.ch'
> Subject: oddity with KS tests
>
>
>
> We're using the TH1::KolmogorovTest function to compare histograms for our code release systems tests. For some fraction of the histograms we get the warning that the histogram is weighted - and then we get what seem like non-sensical KS values.
>
>
>
> I've attached two histogram files. For histogram TKRTRKENERGY, the KS value is (essentially) zero. I've attached the overplot of the histograms.
>
>
>
> In any case, the histograms are *not* weighted, so we're wondering why the KS test thinks they are.
>
>
>
> Here's the snippet of code filling the histogram:
>
>
>
> dummy = new TH1F("TKRTRKENERGY","Track Energy",200, 0., 2000.);
>
>
> Double_t trkE = track->getKalEnergy();
> ((TH1F*)GetObjectPtr("TKRTRKENERGY"))->Fill(trkE);
>
>
>
> This is for Root 3.04.02 on linux RH7.2.
>
>
>
> root [1] TFile* f = new TFile("/nfs/farm/g/glast/u05/systests/Gleam/v5r1p1/Ver$
>
> root [2] TFile* g = new TFile("/nfs/farm/g/glast/u05/systests/Gleam/v5r0p6/Ver$
>
> root [3] TH1* hf2 = f->Get("TKRTRKENERGY")
>
> root [4] TH1* hg2 = g->Get("TKRTRKENERGY")
>
> root [5] double x = hf2->KolmogorovTest((TH1*)hg2)
>
> Warning in <TH1F::KolmogorovTest>: Weighted events and no Sumw2, hist:TKRTRKENER
>
> GY
>
>
>
> Warning in <TH1F::KolmogorovTest>: Weighted events and no Sumw2, hist:TKRTRKENER
>
> GY
>
>
>
> root [6] x
>
> (double)4.16499306271588573e-40
>
>
>
> Regards,
>
>
>
> Richard
>
> --
>
> Richard Dubois
>
> Stanford Linear Accelerator Center
>
> Stanford, CA
>
> 650-926-3824
>
> 650-926-8616 (FAX)
>
> http://www.slac.stanford.edu/~richard
>
>
>
>
This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:10 MET