Plotting Minimized NLL gives different results

[hym]

Dear Experts,
I am doing unbinned maximum likelihood fitting with RooSimultaneous (2 categories), which gives me the following results.

 FCN=-76742 FROM MIGRAD    STATUS=INITIATE       12 CALLS          13 TOTAL
 FIRST CALL TO USER FUNCTION AT NEW START POINT, WITH IFLAG=4.
 START MIGRAD MINIMIZATION.  STRATEGY  1.  CONVERGENCE WHEN EDM .LT. 1.00e-03
                     EDM= unknown      STRATEGY= 1      NO ERROR MATRIX       
  EXT PARAMETER               CURRENT GUESS       STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  fB           1.00000e-01   5.00000e-02   1.72186e-01  -2.89222e+02
   2  fsig_ff      1.00000e-01   5.00000e-02   1.72186e-01   2.22965e+03
   3  fsig_fs      1.00000e-01   5.00000e-02   1.72186e-01   2.11380e+03
                               ERR DEF= 0.5
 MIGRAD MINIMIZATION HAS CONVERGED.
 MIGRAD WILL VERIFY CONVERGENCE AND ERROR MATRIX.
 COVARIANCE MATRIX CALCULATED SUCCESSFULLY
 FCN=-77407.8 FROM MIGRAD    STATUS=CONVERGED     124 CALLS         125 TOTAL
                     EDM=2.53449e-06    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  fB           3.36934e-01   4.44038e-02   1.77798e-02  -4.30100e-03
   2  fsig_ff      2.90215e-02   2.72814e-03   3.12182e-03   9.38526e-02
   3  fsig_fs      1.59796e-02   2.85037e-03   4.30456e-03  -5.30582e-04

But when I try to plot NLL for one category then this value changes and gives the following…

 START MIGRAD MINIMIZATION.  STRATEGY  1.  CONVERGENCE WHEN EDM .LT. 1.00e-03
 FCN=-122430 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  fB           3.36934e-01   4.43149e-02   9.39438e-02   2.98954e+03
   2  fsig_ff      2.90215e-02   2.72715e-03   1.62636e-02  -2.10721e+04
                               ERR DEF= 0.5
 MIGRAD MINIMIZATION HAS CONVERGED.
 MIGRAD WILL VERIFY CONVERGENCE AND ERROR MATRIX.
 COVARIANCE MATRIX CALCULATED SUCCESSFULLY
 FCN=-136092 FROM MIGRAD    STATUS=CONVERGED      42 CALLS          43 TOTAL
                     EDM=6.64737e-05    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  fB           1.90733e-09   3.58337e-05   3.05037e-03** at limit **
   2  fsig_ff      3.25133e-01   2.61684e-03   1.42321e-03  -6.50984e-01

I don’t understand why this situation occurs here, please share your opinions.
For a complete fitting overview. I have attached the full log info here. Output_log.txt (26.8 KB)

[jalopezg]

Hello @hym,

I am not sure I understood your question. I think it would be useful if you give some more information on what you are trying to achieve.
Nevertheless, I am inviting @moneta to this thread so that he can also take a look and shed some light.

Cheers,
J.

[hym]

Hi @jalopezg
Thanks for your concern!
I am have RooSimultaneous pdf with 2 categories. I am performing an unbinned max likelihood fit with this function.

simPdf->fitTo(*data)

and I want to plot negative log likelihood (NLL) for this simultaneous fit with respect to fB.

[moneta]

Hi,

I don’t see any problem looking at the log file. Also the log file does not seem consistent for what you have shown in the posts. If you have a problem, please post a reproducible macro running your fit and your RooFit workspace, so we can run it and see also what the issue is.
Cheers
Lorenzo