Access/Evaluate Likelihood Function

[ilam]

Hi everyone,

I have a question regarding directly accessing the likelihood function in RooFit. I currently have my p.d.fs as RooHistPdf objects and I then use this to fit to some binned dataset. Just for sanity check and out of curiosity, I want to manually fix the values of my parameters and see what value of the likelihood comes out. For example, I want to fix par_1= x and par_2 =y and see what value of FCN I get. In this case, no fitting is being done and is therefore independent of the actual dataset I end up fitting to. Is there a way to do this? Thanks!

[jonas]

Hi! You are currently fitting with RooAbsPdf::fitTo(), right?

You can create the negative log-likelihood function in a similar way with RooAbsPdf::createNLL().

This will return you a RooAbsReal* for the NLL. For your sanity check, you need to set the values of all parameters manually with setVal(), and then you can evaluate the NLL with nll->getVal().

Please let me know if this works for you and if you have further questions!

[ilam]

Hi! Sorry for the slow reply. Yes, I am currently fitting with RooAbsPdf::fitTo(). So you are saying I should do something like this:

RooAbsPdf *abspdf;
RooAbsReal *nll = abspdf->createNLL();
parameter1->setVal(4.0)
parameter2->setVal(3.0)
double c = nll->getVal()

Is this what you meant? Thanks!

[jonas]

Yep, exactly! If you used to pass some arguments to fitTo(), you can also pass the same arguments to createNLL() by the way.

[ilam]

Just came back to say that it worked like a charm! Thank you!