RooMCStudy: pull calculation and getting generated values of parameters

[amroo]

Hello all,

I am doing a toy MC study using RooMCStudy.

I have different generating and fit pdf. For my generating pdf, I am using the Extended() option which I believe samples the generated number of events according to Poisson distribution.

I can plot the pull distributions for all variables without any problems as solved in this post:

Getting pull distribution in RooMCStudy when using separate generator and fit models

similar issue asked here with no reply : 1> Pulls from RooMCStudy with different generating and fitting models 2> RooMCStudy pulls fail using different generate and fit models I am doing a toy MC study for my fitting pdf. I am using different generator and fitting pdf. Both generator and fitting pdf have same shape. The only parameters different in generator and fit pdf are the yields for signal and background. For generator pdf it is fixed and for fit pdf it is floated with some random i…

I want to know the generated value of the number of signal and background for each toy experiment. How can I do that?

I know how to obtain the fitted values of parameters for each toy experiment but no success in doing the same for generated values.

For my fit result I get the following message:
1> The fit parameter ‘n1f’ is not in the model that was used to generate toy data. The parameter ‘n1’=4529 was found at the same position in the generator model. It will be used to compute pulls. (n1 is signal yield). But since I am using the Extended() option, n1 generated should be different for every toy experiment ? The same problem with the background yields.

The relevant piece of code:
RooMCStudy *mcstudy = new RooMCStudy(simPdf,RooArgSet(x,y),Extended(),FitModel(simPdf_f),FitOptions(NumCPU(4),Save(kTRUE),PrintEvalErrors(0),Minos(1)));
mcstudy->generateAndFit(100);

Similar questions asked in this post:

RooMCStudy

Hi, I am trying to get the generated value of “f” as shown in the attachment (filled as h_f_gen)…(below)one.cc (1.33 KB) Is this the correct way to get generated “f” for toy samples ? (2) Having said that i looked at root.cern.ch/root/html/tutorial … str.C.html What is “f_gen” in line-60 ? How does generated “f” is filled in that histogram ? Shouldn’t it have been TH1* h_f_gen = mcs.fitParDataSet().createHistogram(“f”,-40) ; //f instead of TH1* h_f_gen = mcs.fitParDataSet().createHistog…

clarification on the issue will be really helpfull
Thanks

[StephanH]

Hi @amroo,

this information is saved with the fit parameters. The fitParams() or the fitParDataSet()
contain a variable / column that is called “ngen”. It holds the number of events generated in each run.

These could e.g. be plotted like this:

mcs.fitParDataSet().createHistogram(“ngen”, <numberOfBins>) ;

[amroo]

Hi @StephanH Thanks for replying.
yes, I see there is a parameter ngen in fitParDataSet().
ngen is total number of events generated in a run

But how do I get the generated value of other parameters of pdf for which toy study is performed ?
I have something like pdf = Nsig*sig_pdf + Nbkg*bkg_pdf , ngen = Nsig+nbkg.
I want to get the value of nsig and nbkg generated for every run.

Also, what is the definition of pull used here (nsig_fit(i) - nsig_gen(i))/(sigma_nsig_fit(i)) ;
Is the nsig_gen(i) fixed or different for every i while calculating pull ?

[StephanH]

Hi,

unless fit parameters are constrained to a certain interval, the “target” values of the parameters when generating are always the same. Only the target value ngen is changing, all the rest is constant.

Then, events are randomly assigned to the two components with a probability that reflects the relative fractions of nsig and nbkg. The actual number of signal events in each dataset is unknown, but by construction really close to

nsig_gen = ngen * nsig/(nsig+nbkg)

and similar for the background.

The pull is (meas - truth)/delta(meas), and truth is fixed.

[amroo]

Thanks for the detailed clarification. This answers my question.