[Rivet] adding error bars, goodness of fit and sum defects

[Andy Buckley]

Just catching up on this -- somehow I missed this thread before.

As Gavin said, Rivet doesn't do this side of things very accurately: the 
chi2 is just a simple feature in the plotting code we use, and not 
something you should rely on for computing statistics. You can also 
enable the chi2 writing now by adding "PLOT:GofType=chi2" to the 
rivet-mkhistos command-line, rather than hacking analysis .plot files or 
writing your own custom .plot file.

If you want to compute statistics properly, the best way in my opinion 
is to do it yourself via a Python script that uses the YODA 
(http://yoda.hepforge.org/) interface to read the data files, etc. If 
you try that, please give us some feedback on whether there are parts 
that could be made simpler. I have considered adding a chi2 function to 
YODA for some time, but while a simple one is easy (trivial!) to 
calculate, there are so many potential variations (for correlations, 
model error treatment, parallelised calculation, etc.) that I've avoided 
it so far. But I can be convinced...

Andy


On 02/02/16 10:37, Gavin Hesketh wrote:
> Hi Eike,
> you might have to provide more specific information - eg which version
> of rivet, the full commend you use, the error message you get...
>
> But you should also know that rivet does not treat correlations between
> uncertainties, so if you are looking at a systematics limited
> distribution the chi2 will not be fully correct.
>
> cheers,
> gavin
>
> On 02/02/16 01:37, Eike Germann wrote:
>> Hello there :)
>>
>> I've been trying to run rivet analyses with additional parameters,
>> adding error bars to the mc values and getting Chi^2 information
>> displayed. I know that I can activate error bars by specifying --mc-errs
>> in rivet-mkhtml, and I can get the Chi^2 added if I modify the .plot
>> files for the analysis I'm looking at in the local/share/Rivet directory.
>> Trying to add the command for error bars there makes rivet *very*
>> unhappy, and so does using a config file with those parameters for
>> make-plots directly. The instructions for make-plots point to a
>> histogram ID, but I don't know how where to find it?
>>
>> I'd further like to get info on how many data points get rejected for
>> the plots - I'm trying to establish an energy window for pTz and don't
>> know how to get to the data I need to figure out a good cutoff point.
>>
>> Any ideas?
>>
>> Thank you kindly
>> Eike Germann
>>
>>
>> _______________________________________________
>> Rivet mailing list
>> Rivet at projects.hepforge.org
>> https://www.hepforge.org/lists/listinfo/rivet
>>
> _______________________________________________
> Rivet mailing list
> Rivet at projects.hepforge.org
> https://www.hepforge.org/lists/listinfo/rivet


-- 
Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow