RE: Timeout error

From: Joachim Vollaire <joachim.vollaire_at_cern.ch>
Date: Fri, 24 Jun 2011 08:09:32 +0000

Hi

In FLUKA the standard deviation for estimators is calculated using different
batches (of several particle histories) and not individual contribution from
the same primary particle like in some other codes. For this reason, it is
always suggested in order to obtain a reliable standard deviation to run a
minimum of 5 independent batches (more is better). With one single batch
you cannot determine the standard deviation. However you are right in some sense,
if you score 1e6 in 5 x 2e5 or 10 x 1e5 then the estimated statistical error
should be equivalent
Cheers
Joachim

-----Original Message-----
From: owner-fluka-discuss_at_mi.infn.it [mailto:owner-fluka-discuss_at_mi.infn.it=
] On Behalf Of Nicholas Bolibruch
Sent: 23 June 2011 16:53
To: fluka-discuss_at_fluka.org
Subject: Re: Timeout error

Sorry to take this topic on a slight detour, but I feel that I need to
ask this question to improve the results (and confidence) of my
simulations.

My understanding of Monte-Carlo simulation is that if I run a simulation
with 5E6 particles in a single run, it would have the same statistical
result as 5 independent runs at 1E6. Is there something else I'm
missing here?

On Wed, 2011-06-22 at 12:56 -0700, Alberto Fasso' wrote:
> 99% is not an error! It is a flag that an error could not be calculated.
> It needs at least 4 (better 5) independent runs to calculate a standard
> deviation. If the number of runs is less than 4, the FLUKA analysis
> tools print 99%. That number has no meaning, it is only a code for
> "error not calculable"
>
> Alberto
>
>
>
Received on Fri Jun 24 2011 - 11:00:04 CEST

This archive was generated by hypermail 2.2.0 : Fri Jun 24 2011 - 11:00:05 CEST