RE: normalized histograms errors

Hi Ben,

>From quickly running your macro I can't see anything wrong with the
error bars. You have two separate histograms H and K, one with 9 entries in total and one with 3. So when you normalize them, they get of course scaled by different factors. That's why after the normalization, the x=6 bin has a larger error bar rhan the x=3 bin.

To confirm this, put all your entries in a single histogram. Then the errors on both these bins stay exactly the same.

Best,

        Tobi

---
Tobias Raufer
Denys Wilkinson Building
Keble Road, Oxford OX1 3RH, UK

Tel.  +44 (0)1865 273385
Fax. +44 (0)1865 273418
tobias.raufer_at_physics.ox.ac.uk
 



>-----Original Message-----

>From: owner-roottalk_at_pcroot.cern.ch 

>[mailto:owner-roottalk_at_pcroot.cern.ch] On Behalf Of Ben Kilminster

>Sent: 23 November 2005 20:07

>To: roottalk_at_pcroot.cern.ch

>Subject: [ROOT] normalized histograms errors

>

>Hi,

>

>The errors on my normalized histograms are not what I'd expect 

>them to be.  To demonstrate this, I have included a macro (see below).

>

>In the following macro, I plot two histograms on the same 

>canvas.  The top histogram has two histograms overlayed.  The 

>bottom canvas is the 

>same but with the histograms normalized.   Notice that in the bottom 

>canvas the errors are larger for the x=6 bin than for the x=3 

>bin even though they have the same number of entries, and have 

>the same-sized error bars in the top canvas.

>

>Is this the desired behavior ?

>If so, is there a way I can switch the errors of normalized 

>histograms so that their errors are relatively comparable  

>(Ie, error in bin 3 is the same as the error in bin 6) ?

>I have tried using TH1F::Sumw2() before plotting them in the 

>second plot, but no luck.

>

>~Ben

>

>{

>

>TH1F *K = new TH1F("K","No norm",10,0,10); TH1F *H = new 

>TH1F("H","Norm",10,0,10);

>K->SetLineColor(kBlue);

>

>H->Fill(2);H->Fill(2);H->Fill(2);H->Fill(2);H->Fill(2);H->Fill(2);

>H->Fill(3);H->Fill(3);H->Fill(3);

>

>K->Fill(6);K->Fill(6);K->Fill(6);

>

>TCanvas *C = new TCanvas("C", "test", 600,800);

>C->Divide(1,2);

>// top canvas has no normalization

>C->cd(1);

>H->DrawCopy("elp");

>K->DrawCopy("same, elp");

>

>H->SetNormFactor();

>K->SetNormFactor();

>

>// bottom canvas is normalized

>C->cd(2);

>K->Draw("elp");

>H->Draw("same, elp");

>

>}

>

>