WHIZARD

rescale weight

Asked by Si Wang

Hi,

I saw this message in log output:
 Dropped events (weight zero) = 225101 (total 1225101)
Warning: All event weights must be rescaled by f = 8.16259231E-01

Then what to do?
In python analyze weighted LHE file, to plot differential distribution,
I use hist=np.histogram(costh,bins=20,range=(-1.0,1.0),weights=weight2)
Then normalize hist by times luminosity and cross section, then divide the sum of weights.
The result seems not correct.

I guess this warning may have an influence.
Where to times this rescale factor f?

Thank you!
Si

Question information

Language:
English Edit question
Status:
Solved
For:
WHIZARD Edit question
Assignee:
Pia Bredt Edit question
Last query:
Last reply:
Revision history for this message
Pia Bredt (bredtpia) said : 		#1

Hi Si,

This Warning is triggered if the statement "?keep_failed_events=false" which is the default. That means if you have set cuts and you have a target fixed number of events you want to have, events are discarded and produced new unless these events pass the cuts. Your event file in the end thus consists of events which indeed passed the cuts and the total number of produced events (including those which failed the cuts) exceeds your target number of events which are in your event file. This in general can give you a wrong normalization of distributions which can be recovered by multiplying the event weights (or distributions) by hand by the correction factor

f = <target number of events>/<generated events incl. failed events>

Another way to recover the correct normalization is to use "?keep_failed_events=true" which keeps track of cut-failed events but assign zero weights to those. This of course gives you less statistics, but no factor f is needed in this case.

Kind regards,
Pia

Revision history for this message
Juergen Reuter (j.r.reuter) said : 		#2

Changing the status to solved.