Comparision Roostats tool with simple ROOT macro

Roofit and RooStats
1

Dear all,

I wanted to do a basic example of rejection of the background-only hypothese with one bin.

Hence I have an estimation of the background yield “nb”, and an observation of “nobs” events.

What is the significance of the deviation (nobs-nb) ?

I compared a tool provided by RooStats, with a log likelihood ratio calculation with Poisson distribution in a simple ROOT macro.

You can find them there:
iktp.tu-dresden.de/~prudent/ATLA … gnifStat.C
Just make

root -l
.L signifStat.C
plot()

to see the comparison for nb and nobs varying from 0 to 50.
iktp.tu-dresden.de/~prudent/ATLA … _macro.gif

As you can see there are clear differences between the two plots. I must have misunderstood a point, do you see what ?

Thanks in advance,
cheers,
Xavier

2

Hi,
These are two different methods for computing the significance, see Cousins,Linnemann and Tucker paper:
arxiv.org/pdf/physics/0702156v4

The first one is Z_BI (see paragraph 3) , while the other one is the profiled likelihood method (see par. 5)
You can compute the Z_PL also in RooStats using the ProfileLikelihood calculator

Best Regards

Lorenzo