HistFactory inter- and extrapolation defaults reference

Roofit and RooStats
1

I was trying to find out what the default inter- and extrapolations were for OverallSys and HistoSys in ROOT 6 (specifically Root 6.08).

In the original reference (https://cds.cern.ch/record/1456844/files/CERN-OPEN-2012-016.pdf) it states that code 0 (piece linear) is the default for HistoSys and 1 (piece exponential) for OverallSys.

However i see in master documentation (ROOT: PiecewiseInterpolation Class Reference) that now code 4 (poly inter and linear exp) is the default for HistoSys.

However i could not find what the default for OverallSys is currently and if both of those are true for Root 6.08 as there it is not explicitely stated that 4 is the default (ROOT: PiecewiseInterpolation Class Reference)

Is there any reference to looks this up or a way to verify that in the code?

2

Hi,
As far as I remember also in older version like 6.08 the interpolation code used is 4. This was set when creating the HistFactory workspace, see line 514 in ROOT: roofit/histfactory/src/HistoToWorkspaceFactoryFast.cxx Source File

Lorenzo

3

Thanks, I will try to have a look at that.

How does it work for the overallsys? What is the default and where could I reference that?

4

Hi,
The same is happening for the FlexibleInterpVar, see line 728 in the same linked file

Lorenzo

5

Hi,

thanks again. So it would effectively be these links:
https://root.cern/doc/v608/HistoToWorkspaceFactoryFast_8cxx_source.html#l00514
https://root.cern/doc/v608/PiecewiseInterpolation_8cxx_source.html#l00225

https://root.cern/doc/v608/HistoToWorkspaceFactoryFast_8cxx_source.html#l00728
https://root.cern/doc/v608/FlexibleInterpVar_8cxx_source.html#l00453

Where the comment for the // LM: change to 4 (piece-wise linear to 6th order polynomial interpolation + linear extrapolation ) for the default value of FlexibleInterpVar is not accurate as it does exponentiol extrapolation as seen in the respective source code.