[Halld-pid-upgrade] Fwd: Re: thoughts after PID session [from Mike]

[Michael Williams]

Agreed.  Step 1: Find S and B.  Step 2: Convert this into an FOM which is closely related to what we actually want to do.    Even in discovery experiments you often find that the purely stat-based FOMs neglect that you don't know the bkgd PDF and so you'd actually rather have less bkgd than is statistically optimal.  

Anyway, for now let's try and complete Step 1 for more channels.   At MIT I think our approach should be to try and use Paul's code and Pythia sample to do "anything".   Once we can do anything we can then do something useful.    

Mike

On Oct 9, 2012, at 10:53 AM, Matthew Shepherd wrote:

> 
> On Oct 9, 2012, at 7:49 AM, Elton Smith wrote:
> 
>> I agree with Matt that as far as the PID system is concerned, one should 
>> focus on S and B. I also agree that S/B is not the right figure of merit 
>> and we should use something like S/sqrt(S+B).
> 
> 
> Just for the record, I wasn't arguing that S/sqrt(S+B) was superior to S/B.  As Mike pointed out S/sqrt(S+B) is used to optimize signal significance.
> 
> The problem is that backgrounds in amplitude analysis are very difficult to handle.  I'd be in favor of making the background as small as possible even if it resulted in a lower efficiency.
> 
> Another way to say it is that we're going to be driven by systematic errors, so constructing an FOM based on statistical arguments may not be the best.
> 
> The important thing now is developing a method for producing the numbers S and B and then we can discuss how to combine them into an FOM.
> 
> Matt
> 
> 
> 
> _______________________________________________
> Halld-pid-upgrade mailing list
> Halld-pid-upgrade at jlab.org
> https://mailman.jlab.org/mailman/listinfo/halld-pid-upgrade