Re: statistics question ?

[Steven Clayton <smclayto@npl.uiuc.edu>]

On Tue, Apr 11, 2006 at 12:16:00PM -0700, Tom Banks wrote:
> 
> > We have a small (5x5 mm scintillator ) detector
> > whcih we put in the center of the chambers
> > in the place of the TPC.
> >
> > Do you have any suggestions how much of statistics
> > one would need in terms of muon stops to
> > extract the e-tracking resolution on a reasonable level?
> 
> 
> It's difficult to give an exact answer to a vague question, but I wouldn't
> think that you'd need very much... maybe something between 1e6 and 1e7
> events (a couple hours' worth?), although you could probably make do with
> less.  Of course, I'm sure your considerations are constrained by
> experimental exigencies such as the maximum mustop rate, etc.
> 
> I think Steve already performed this study a long time ago (2001?), using
> ePC1 and the functioning quadrant of the damaged ePC2.  Perhaps he could
> give a better answer.
> 
> Tom

I think we want to see the tail region of the impact parameter
distribution well enough.  With 1 mm binning, the amplitude of
the normalized distribution is ~0.001 at b = 100 mm.
1e6 statistics would give ~1e3 in that bin.  If we select only
a particular gondola, then that cuts it by a factor or 16, and
the counts are marginal.  I guess I would vote for 1e7 for each
eye position, if you have the time, or whatever you can get.  The
histogram could be rebinned if there are not quite enough counts
in the tail.

Steven