Problem in using Conditional Observable in 2D fit

[cpkar]

Hi Experts,

I am doing a 2D fit between mass and decay time with per event decay time error. I looked the roofit tutorial “$ROOTSYS/tutorials/roofit/rf306_condpereventerrors.C” to use the perevent error in my fit. First I generate the per event error distribution from pdf and used that to 2D fit. I found that some of my parameters are giving weird value(hitting the limit) after the fit. I don’t know why it differs much from the generated value. Could somebody please point out the reason of getting bias?

An example is attached below.
massfit_toy_test.cc (10.3 KB)

Regards
Chandi

[Axel]

@moneta can you help here?

[cpkar]

Hi All,
Any response?

Thanks in advance
Chandi

[couet]

@moneta is not here right now. Meanwhile I ran your macro. It gives me:

RooFit v3.60 -- Developed by Wouter Verkerke and David Kirkby 
                Copyright (C) 2000-2013 NIKHEF, University of California & Stanford University
                All rights reserved, please read http://roofit.sourceforge.net/license.txt

Active Message streams
[0] MinLevel = PROGRESS Topic =  Any 
size for fitting and seed 
RooDataSet::TotPdfData[m,trecoxy,trecoxye] = 26941 entries

  RooFitResult: minimized FCN value: -255857, estimated distance to minimum: 0.000210171
                covariance matrix quality: Full, accurate covariance matrix
                Status : MINIMIZE=0 HESSE=0 

    Constant Parameter    Value     
  --------------------  ------------
               meanGct    0.0000e+00
                 meant    0.0000e+00
                sigmat    1.0000e+00

    Floating Parameter  InitialValue    FinalValue +/-  Error     GblCorr.
  --------------------  ------------  --------------------------  --------
                  Nsig    2.1434e+04    2.4176e+04 +/-  2.20e+02  <none>
            Tautotalsb    1.3000e+00    1.3102e+00 +/-  1.52e-01  <none>
                   div    2.4000e-02    1.0000e+00 +/-  6.98e-01  <none>
                  fbkg    3.8862e-01    4.2990e-01 +/-  5.09e-02  <none>
                   fg1    1.0613e-01    8.6187e-02 +/-  8.62e-03  <none>
            fracdecay2    3.2578e-01    2.9863e-01 +/-  1.22e-01  <none>
                lambda   -1.6890e+00   -9.8645e+00 +/-  1.37e+00  <none>
                  mean    5.2740e+00    5.2742e+00 +/-  2.08e-04  <none>
                  nbkg    5.5045e+03    2.7650e+03 +/-  1.64e+02  <none>
              shoulder    5.1357e+00    5.2500e+00 +/-  1.92e-01  <none>
                sigma1    8.0160e-02    8.7489e-02 +/-  6.69e-03  <none>
                sigma2    2.7179e-02    2.7731e-02 +/-  2.28e-04  <none>
              sigmaGct    2.3000e+00    2.2571e+00 +/-  4.43e-01  <none>
                     t    1.6160e+00    1.6064e+00 +/-  1.28e-02  <none>

 FCN=147.707 FROM MIGRAD    STATUS=CONVERGED      19 CALLS          20 TOTAL
                     EDM=8.1909e-10    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  p0          -2.13547e-02   9.54491e-02   5.68338e-04  -4.24042e-04
 FCN=92.4933 FROM MIGRAD    STATUS=CONVERGED      15 CALLS          16 TOTAL
                     EDM=6.58516e-15    STRATEGY= 1      ERROR MATRIX ACCURATE 
  EXT PARAMETER                                   STEP         FIRST   
  NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE 
   1  p0           7.24028e-03   1.10432e-01   5.21379e-04  -1.03921e-06

and the following picture:

What do you think is wrong with these outputs ?

[cpkar]

Hi,

Thanks for looking into this. Have you seen the mass distribution?

div         2.4000e-02    1.0000e+00 +/-  6.98e-01  <none>
shoulder    5.1357e+00    5.2500e+00 +/-  1.92e-01  <none>

These two variables are hitting the limit which can be seen in the mass plot.

Thanks
Chandi

[couet]

Ok. I Iet @moneta have a look when he will be back.