Messy output of error matrix and correlation matrix

Roofit and RooStats
1

Hi all,

Since I’m using MINUIT, the output of the error matrix and the covariance matrix get messy when the number of fit parameters exceeds a certain limit. I’m using MINUIT only through the RooFit interface, if this is a matter. For some reason lines get replicated partially. Since this doesn’t get fixed, is there a deeper reason for it?

The printout of the correlation matrix gets so messy I believe not even the full matrix gets printed (I spare you the output).

 EXTERNAL ERROR MATRIX.    NDIM=  73    NPAR= 36    ERR DEF=0.5
 ELEMENTS ABOVE DIAGONAL ARE NOT PRINTED.
  3.633e-03 
 -2.500e-05  2.176e-04 
 -1.016e-05  2.107e-07  3.119e-05 
  1.187e-07  5.549e-08  1.456e-07  4.340e-07 
 -1.603e-07 -4.507e-10  1.564e-07  1.303e-09  6.671e-07 
 -1.487e-07  2.670e-08  1.175e-07  1.078e-09  1.239e-09  1.426e-07 
 -2.208e-07  2.792e-08 -2.949e-08 -4.436e-08 -4.315e-10 -2.413e-10  2.835e-07 
  9.524e-09  9.345e-09 -4.570e-08 -4.187e-10 -4.737e-08 -4.416e-10  8.490e-10  4.215e-07 
 -7.143e-08 -2.152e-08  4.187e-09 -2.589e-11 -2.163e-10 -8.639e-09  7.768e-10  8.416e-10  9.150e-08 
 -1.644e-06  8.301e-07 -1.743e-05 -2.534e-07 -3.928e-07 -2.760e-07  4.373e-07  5.178e-07  4.327e-07  2.830e-04 
  1.206e-06 -3.684e-07 -2.520e-06 -5.178e-08 -4.546e-08 -5.015e-08 -5.088e-08 -4.467e-08 -6.354e-08 -1.752e-05  8.756e-06 
 -7.012e-02 -9.348e-05 -3.909e-03  8.989e-06  2.960e-05  3.303e-05 -1.363e-04  4.380e-06  1.470e-05 -4.282e-03 -4.551e-03 
 -7.012e-02 -9.348e-05 -3.909e-03  8.989e-06  2.960e-05  3.303e-05 -1.363e-04  4.380e-06  1.470e-05 -4.282e-03 -4.551e-03  5.252e+02 
  8.484e-02  5.119e-04 -4.886e-03  9.947e-05  2.217e-05  2.625e-05 -2.314e-04  1.847e-06  7.980e-06 -6.129e-03 -4.289e-03 
  8.484e-02  5.119e-04 -4.886e-03  9.947e-05  2.217e-05  2.625e-05 -2.314e-04  1.847e-06  7.980e-06 -6.129e-03 -4.289e-03  3.795e+00  5.301e+02 
 -1.786e-01 -6.573e-04 -1.845e-03  1.751e-04  2.514e-04  2.980e-04 -1.146e-04 -1.781e-04 -3.149e-04 -1.171e-01 -8.112e-03 
 -1.786e-01 -6.573e-04 -1.845e-03  1.751e-04  2.514e-04  2.980e-04 -1.146e-04 -1.781e-04 -3.149e-04 -1.171e-01 -8.112e-03  1.871e+01  1.197e+01  1.026e+03
2

You can store this matrix into a TMatrixD if you like

Rene

3

Hi,

The full correlation matrix information can also be retrieved from
the RooFitResult object. To get a RooFitResult, save the return value
of RooAbsPdf::fitTo() and do not forget to specify the Save() option
in fitTo() to request that the fit result object should be created.

Wouter

4

Rene, Wouter,

Thanks for your answers. So no need to get worried. Only I’m wondering why nobody took the time to fix the printout. On the other hand, I wouldn’t do it either :slight_smile:

  • Moritz