Extracting Maximum Likelihood from a Fit

Roofit and RooStats
1

I am fitting two different user-defined functions (TF1 *f & *g) seperately to a histogram (TH1F *h) using the likelihood method: h->Fit(“f (or g)”, “L”);

The fits converge and everything is good, but my ultimate goal is to perform a likelihood ratio test between the two models that the functions describe. Is there a simple way to extract the maximum likelihood of each model from the fit results?

2

Hi,

You can get the likelihood value from the TFitResult object which is obtained when using option “S”.
Here is the code example:

TFitResultPtr r = h->Fit( f, "L S");
double likelihood_value = r->MinFcnValue();

Best Regards

Lorenzo

3

Lorenzo,

I am concerned by the name “MinFcnValue”. The likelihood should be a maximum. Is the term “min” used because r->MinFcnValue() may return chi2 (a minimum) or likelihood depending on the fit option?

(I just want to make sure I’m getting the right number.)