Linear Collider Forum



Home » Analysis and Reconstruction » Analysis Tools » Efficiency plot using AIDA
Re: Efficiency plot using AIDA [message #1548 is a reply to message #1510] Fri, 11 July 2008 11:51 Go to previous messageGo to previous message
partridge
Messages: 11
Registered: October 2005
Hi Tony,

It looks like the code from Adam is calculating 95% CL intervals instead of 1 sigma error bars. That probably explains why the errors were a factor of ~2 too big compared to my estimates using binomial errors.

When there are more than 10 entries, this is explicit, as shown from the code fragment from EfficiencyBinError shown below:

69: if (NX >= 10) {
70: // NUMBERS ARE ALL LARGE, SO CAN USE ANALYTIC CALCULATION
71: return 1.96*Math.sqrt(R1*(R2-R1)/R2)/R2;
72: }

What is being returned is just a factor of 1.96 times the binomial error.

I personally don't think 95% CL errors are all that useful - when people look at a plot they are expecting 68% CL errors unless there is a good reason for doing something different.

I also am not sure exactly what Adam was doing for the case of less than 10 entries - it looked like there were a bunch of special cases and it is hard to tell by looking at the code if they make sense.

I would be happy to code a version of EfficiencyBinError that simply returned binomial errors if that would be useful.
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic:JAS3 classpath question/trying to get trivalPFA to compile
Next Topic:Fitting histograms with other histograms
Goto Forum:
  


Current Time: Fri Apr 20 04:01:02 Pacific Daylight Time 2018
.:: Contact :: Home ::.

Powered by: FUDforum 3.0.1.
Copyright ©2001-2010 FUDforum Bulletin Board Software