Sumw2 method for weighted histogram

[trenta_coollime]

---

ROOT Version: 6.06/01
Platform: Ubuntu 16.04
Compiler: gcc530

---

Hello, root experts.

I have a question for usage of Sumw2 function in my situation.

At the moment, I am filling my histogram in given way

sample ='ZH_HToSSTobbbb_ZToLL_MH-125_MS-15_ctauS-10 & 5'
h16 = TH1F("Decaydist",'%s Decaydist' % sample,100, 0, 100)
h16w = TH1F("Decaydistw",'%s Decaydistw' % sample,100, 0, 100)
h16w.Sumw2()
h16s = TH1F("ctauEventWeight",'%s ctauEventWeight' % sample,80, 0, 4)
DY16dir=glob.glob('path/0000/*')
for i in range (7):
        filename = DY16dir[i]
        _file0 = TFile.Open(filename,'read')
        mytree16 = _file0.Get("lldjNtuple/EventTree")
        for entry in mytree16:
                ctau = mytree16.ctauEventWeight
                h16s.Fill(ctau[0],1)
                #print(ctau[0]) 
                Decaydist=mytree16.AODnVtx
                h16.Fill(Decaydist,1)
                h16w.Fill(Decaydist,ctau[0])

As you can see, h16w hist is weighted with ctau[0] with predefined condition of h16w.Sumw2()

At the end, the plot looks like

Technically, the integral of the red(h16) and the blue(h16w) should be equal to each other, but differ a little bit in my case.

I want to prove that the discrepancy in integral is OK because the difference is within the error range of the integral of the Weighted (blue, h16w) histogram".

How can I get the error range in the integral of the blue histogram, using the Sumw2?
Do you have any good idea?

Thank you.

Trentacoollime

[Wile_E_Coyote]

In general, both integrals would be exactly equal only if for all entries “ctau[0]=1”.

Try:

h16w.Fill(Decaydist, ctau[0] / 10.)

[trenta_coollime]

I have a reason why it should be equal to 1…

When I do h16w.Fill(Decaydist, ctau[0] / 10.),

[Wile_E_Coyote]

So, as you can see, everything seems to be fine (if weights are 10 times smaller, so is the integral).
You could also try:

print(h16w.Integral() - h16.Integral())

And if, for test purposes, you want them exactly equal, try:

h16w.Fill(Decaydist, 1)

[trenta_coollime]

Filling the histogram indeed seems all fine.

Yet, I want to know the statistical error of my h16w histogram’s integral using the Sumw2().

I do set h16w.Sumw2() before filling the hist, but how do I retrieve the information of statistical error with Sumw2()?

Thanks!

[Wile_E_Coyote]

TH1::IntegralAndError