[Roofit] Why is NLL minus?


I made pdf for Relative Breit-Wigner using RooClassFactory::makePdf().
Then I have tested it like this.

    using namespace RooFit;

    RooRealVar x("x", "x", 1, 3);
    RooRealVar mass("mass","mass",1.530, 1.0, 3.0);
    RooRealVar width("width","width",0.336, 0.2, 0.4);
    RooRealVar mdau1("mdau1", "mdau1", 0.9382720);
    RooRealVar mdau2("mdau2", "mdau2", 0.139570);
    RooRealVar Lmin("Lmin", "Lmin", 1);
    RooRealVar hp("hp", "hp", 0);
    RooRelBW relBW("relBW", "relative Breit-Wigner", x, mass, width, mdau1, mdau2, Lmin, hp);

    RooAbsData* data = relBW.generate(x, 10000);

    RooFitResult* r = relBW.fitTo(*data, Save());
    RooPlot* xframe = x.frame();
    relBW.paramOn(xframe, Format("NEU", AutoPrecision(4)));
 FCN=-1260.83 FROM MINOS     STATUS=SUCCESSFUL     39 CALLS          80 TOTAL
                     EDM=2.233e-05    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                  PARABOLIC         MINOS ERRORS        
  NO.   NAME      VALUE            ERROR      NEGATIVE      POSITIVE   
   1  mass         1.53006e+00   2.03063e-03  -2.02618e-03   2.03544e-03
   2  width        3.30644e-01   5.53109e-03  -5.45690e-03   5.61460e-03
                               ERR DEF= 0.5

Why is NLL(FCN) minus?

P.S. Are there some parameters to judge the fit quality like chi^2/ndf and C.L. in mn_fit?

RooRelBW.cxx (3 KB)
RooRelBW.h (1.69 KB)