[Rivet] yodamerge for Histo1D

[Leif Lönnblad]

On 2013-06-26 11:30, Frank Siegert wrote:
> I was trying to understand the merging of d\sigma/dX-type histograms,
> and got stuck in the "For each bin..." paragraph on page 1. I assume
> there must be a sum over i missing in \sigma[b] and \delta\sigma[b]?

Yes. Or rather that the subscript i should not be there:
$\sigma^{[b]}=\sigma S_{w}^{[b]}/S_{w}$

> Anyway, in the end what I'm after, as far as I can tell, are the two
> equations at the end of page 1 and beginning of page 2. The problem
> here, I think, is that we don't have all the information necessary:
> the total sum of weights in run i (S_{w,i}) isn't stored anywhere, is
> it?

Yes, and no. I think most analyses actually need to know the total sum 
of weights and the total cross section for the run, so it is probably 
stored somewhere. But the information is probably discarded after 
finalize(). I think Rivet should centrally store the sum of weights and 
the cross section for each run.

To be completely general, I also think that both the raw histograms and 
the finalized version should be stored, and that after the merging of 
files, it should be possible to run finalize() again.

/Leif