Migrad first derivative ** at limit **

Roofit and RooStats
1

Dear Experts,

I am running an unbinned likelihood fit on a data sample with N_bkg and N_sig as the parameters. I define the variables with limits as

  RooRealVar nsig("nsig","Number of Events",1e4,0,1e9);
  RooRealVar nbkg("nbkg","nbkg",7.8e7,1e2,1e10);

and the result gives me

 START MIGRAD MINIMIZATION.  STRATEGY  1.  CONVERGENCE WHEN EDM .LT. 1.00e-03
 FCN=3248.94 FROM MIGRAD    STATUS=INITIATE        8 CALLS           9 TOTAL
                     EDM= unknown      STRATEGY= 1      NO ERROR MATRIX       
  EXT PARAMETER               CURRENT GUESS       STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  nbkg         7.80000e+07   3.90000e+07   4.58879e-02   4.90198e+05
   2  nsig         1.00000e+04   5.00000e+03   1.63692e-03   2.99360e+06
                               ERR DEF= 0.5
 MIGRAD MINIMIZATION HAS CONVERGED.
 MIGRAD WILL VERIFY CONVERGENCE AND ERROR MATRIX.
 COVARIANCE MATRIX CALCULATED SUCCESSFULLY
 FCN=-36149.9 FROM MIGRAD    STATUS=CONVERGED      46 CALLS          47 TOTAL
                     EDM=1.60367e-06    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  nbkg         9.50256e+06   1.23673e+05   5.25488e-05   3.08616e+00
   2  nsig         1.91875e+02   1.73435e+01   5.18292e-06** at limit **
                               ERR DEF= 0.5
 EXTERNAL ERROR MATRIX.    NDIM=  25    NPAR=  2    ERR DEF=0.5
  1.529e+10 -1.643e+05 
 -1.643e+05  3.008e+02 
 PARAMETER  CORRELATION COEFFICIENTS  
       NO.  GLOBAL      1      2
        1  0.07662   1.000 -0.077
        2  0.07662  -0.077  1.000
now running minos: 
 **********
 **   18 **MINOS        1000           2
 **********
 FCN=-36149.9 FROM MINOS     STATUS=SUCCESSFUL     24 CALLS          71 TOTAL
                     EDM=1.60367e-06    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                  PARABOLIC         MINOS ERRORS        
  NO.   NAME      VALUE            ERROR      NEGATIVE      POSITIVE   
   1  nbkg         9.50256e+06   1.23673e+05                            
   2  nsig         1.91875e+02   1.73435e+01  -1.69150e+01   1.77759e+01
                                 WARNING -   - ABOVE PARAMETER IS AT LIMIT.
                               ERR DEF= 0.5

So I’m warned I’m at a limit, but my lower limit is 0, not 1.91875e+02. Am I truly on a boundary?

2

Hi,

from the Minuit log yes it looks like your parameter is at the limit. Are you sure you do not changes the limits afterwards ?
To be sure, just print the parameter limits before fitting,

Lorenzo