Re: Low Levels Environmental -REPLY

[RADPROJECT@aol.com]

In a message dated 9/30/99 10:59:28 AM Eastern Daylight Time, 
HoodaB@wipp.carlsbad.nm.us writes:

<< Subj:     Low Levels Environmental
 Date:  9/30/99 10:59:28 AM Eastern Daylight Time
 From:  HoodaB@wipp.carlsbad.nm.us (Hooda, Benny)
 Sender:    radsafe@romulus.ehs.uiuc.edu
 Reply-to:  <A 
HREF="mailto:radsafe@romulus.ehs.uiuc.edu";>radsafe@romulus.ehs.uiuc.edu</A>
 To:    radsafe@romulus.ehs.uiuc.edu (Multiple recipients of list)
 
 Hi,
 I need peer opinion on very low levels of environmental activity where the
 results reported are mostly negative concentrations. The DOE/EH 0173T,
 "Environmental Regulatory Guide for Radiological Effluent Monitoring and
 Environmental Surveillance" , section 7.3.4, states that less- than-
 detectable values and negative values have to be included in the data
 evaluations, and assigning zero, detection limit, or some in between value
 would "severely" bias the resulting estimates and should be avoided.
 Therefore, I have been including these negative values in the statistical
 analyses. However, I am being asked to use the MDA instead of negative
 values. What's your input? >>
================
Dear Radsafer:
You mention you are being asked to use "MDA instead of negative values". 
Whose asking you?

Every analytical measurement should have a reported analytical value +/- its 
uncertainty value [ x +/- y] to be meaningful in deriving averages and 
background concentrations. The analytical value, x,  reported can be 
negative, 0, or positive.

This is quite important since by including negative values based on your 
measurement you get to determine if there is an analytical bias to the 
measurment process. For example, in the case where there is a "known" or 
likely zero concentration of some isotope such as I-131 [say in a pre-op 
survey, during a period of no nuclear fallout from any source] a series of  
regular measurements should yield a frequency distribution about zero 
measured activity. Some measurments should be negative, some positive, most 
near zero. If you plotted all your measurments out as a histogram you would 
hopefully see a sort of bell shaped curve distributed about zero with fewer 
large postive and negative values than measurements closer to zero.

If you saw a distribution that had a mean sizable negative value, something 
wrong wrong with your analytical process [incorrect background or bias of 
some type]. If you saw a distribution which had a positive mean, again 
something is likely wrong with your measurment process [assuming no real 
positive backgound activity for the isotope in question]

If you instead listed actual zero or negative measurments as some calculated 
MDA value you would be biasing your calculated average for the isotope in 
question to the high side. This is not something you would want to do for a 
many reasons.

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