[ RadSafe ] Validity / Usefulness of a result....

[Brennan, Mike (DOH)]

Hi, Pete.

There are several things that need to be considered before you come up
with guidelines, because they will effect the answers.

1.  Why are you sampling?  If the reason is to show that the
concentrations and/or totals in the tank does not exceed some limit that
you are no where close to, then it almost doesn't matter what guidelines
you choose.  If, on the other hand, this is a critical quality control
sampling, then you may need to look at changing your
hardware/software/sampling/analysis combo to lower your MDA and
uncertainty.  

2.  Assuming that you've decided that your MDA and uncertainty are
acceptable, you have to look at your sampling frequency compared to what
is going on in the tank.  If you are sampling quarterly, but the
contents of the tank turn over daily, then it doesn't matter, as you are
fulfilling a poorly conceived regulatory function, rather than doing
something real.  If, on the other hand, decisions about what goes into
or out of the tank are based on your sample results, then the frequency
of sampling must be great enough to capture the process you are trying
to control.  Guidelines that might result in a number of samples not
included may not me acceptable.

3.  If your sample frequency is large compared to the cycling of the
tank, I would say just use all your results.  Just be aware that you do
not have great confidence in them.

In my piece of the rad bizz we collect environmental samples, often with
an eye on showing that they are not contaminated.  As one would expect,
most of the time we don't find the isotopes of concern.  We generally
conclude that we have not found evidence that the isotope is in the
sample if the result is less than the MDA or less than the uncertainty.
Although we usually report the numbers, we consider these to be
non-detect results. 

-----Original Message-----
From: radsafe-bounces at radlab.nl [mailto:radsafe-bounces at radlab.nl] On
Behalf Of Pete_Bailey at fpl.com
Sent: Tuesday, May 01, 2007 6:30 AM
To: radsafe at radlab.nl
Subject: [ RadSafe ] Validity / Usefulness of a result....


Consider a 'periodic sampling of a tank' situation.
Often, the results are quite low, not much over MDA.

Attempting to set up an extremely simple guideline for 'the least
experienced / least trained'
lab person to deal with....

The process I'm attempting to adress:
      ( sample resuts are R +- r ,  r is 2 sigma)
   two results,  R1 =-r1   and  R2 +- r2

Determine if the difference between R1 and R2 is 'real' . . .

The 'statistics book' stuff is easy
     (can get tangled in square root of sum of squares...)

But I'm Thinking of testing any one result for 'confidence'
before doing the math.

I'm thinking of situations where subsequent results are 'kinda
close'....

    IF  R2/r2 exceeds 50%, the result shouldn't be used;
       wait untill next sample and compare R1 and R3

   IF  R2/r2 = or exceeds 25% (but < 50%), can compare
      R1 and R2, but should wait until R3 is available
      to draw firm (actionable) conclusions .

    IF R/r is less than 25 %, then go for it....

OR should I let them use the result, no matter how big  r  is, and let
the test of differences 'sort it out'  ?

Also, I'm looking for support of the 'break points' (25% , 50%) I
'choose'.

TIA

   Pete

_______________________________________________
You are currently subscribed to the RadSafe mailing list

Before posting a message to RadSafe be sure to have read and understood
the RadSafe rules. These can be found at:
http://radlab.nl/radsafe/radsaferules.html

For information on how to subscribe or unsubscribe and other settings
visit: http://radlab.nl/radsafe/