Re: Chi-Square Requirements

["Kevin Wright" <kswusa@hotmail.com>]

Negating procedural or license requirements, there is no definitive 
requirement.  The Chi-squared Test is a tool to evaluate a system's 
performance.  As with any statistical test it has false positives and false 
negatives. A Chi-squared evaluation on a series of data can be used to 
evaluate a counting systems performance.  It can be used to determine if a 
counting geometry is not reproducable (if properly performed).

A standard practice is to perform a Chi-squared evaluation of a series of 
data immediately after calibration.  This not only evaluates the 
distribution of the data, but allows you to calculate the standard deviation 
of the data to establish control charts.  Typically, 2 and 3 sigma values 
are established.  If you don't do the Chi-squared test, how do you establish 
your chart limits?  There are other ways, but I personnally prefer the 
Chi-squared Test.

After initially establishing your charts, review data as it is collected 
when performing your routine source checks.  Do not be surprised if the data 
does not vary about the mean.  If your Chi-squared was collected on one day, 
environmental or other factors may be effecting the data from routine checks 
that was not accounted for in the initial test.  After collecting the 
desired number of daily counts (typically between 20 and 60), re-perform the 
Chi-squared Test on that data if the data does not vary randomly about your 
original mean.  Hopefully, this new chart will last until your system is 
re-calibrated, or some other system modification causes you to replace it.

Do not fall into the trap of routinely performing a Chi-squared Test and 
re-establishing control charts.  If you have to do this, your system may not 
be operating properly in the environment it is in, regardless of the results 
of each new test.  External factors or slow developing errors may be 
affecting your system. The best control chart is an old control chart!  You 
can evaluate trends in the data being collected over time to determine your 
system's performance.  Trends in control charts should be evaluated to 
identify emerging problems with your counting systems.

A Chi-squared Test is appropriate for counting systems with reproducible 
geometries that have determined efficiencies.

Hope this helps.

Kevin Wright

>From: "Jenkins, Ken A." <KAJENKIN@southernco.com>
>Reply-To: radsafe@romulus.ehs.uiuc.edu
>To: Multiple recipients of list <radsafe@romulus.ehs.uiuc.edu>
>Subject: Chi-Square Requirements
>Date: Tue, 9 Jan 2001 07:40:59 -0600 (CST)
>
>Good morning,
>Can someone point me to guidance on when a chi-square should be performed 
>on
>tabletop counters such as the MS-3 with a GM probe?  Specifically, must a
>chi-square be performed before the instrument is placed in service 
>following
>a repair or calibration?
>
>Thanks for your assistance,
>Ken Jenkins
>Southern Nuclear Operating Co.
>************************************************************************
>The RADSAFE Frequently Asked Questions list, archives and subscription
>information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com

************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html