Jerry brought up some fine practical points as he always does, but I think
the type of data requested, regardless of sample handling protocol, will
contain more "noise" than "signal" for the purpose described.

First, dead bees collected from the front of a hive after a typical
pesticide kill may include some number of bees that died of "natural
causes", and were tossed out of the hive into the pile of pesticide-killed
bees.  As pesticide screenings are done on groups of bees, significantly
lower (apparent) values will result if one or more of the bees in the sample
died of natural causes, yielding misleading data.

So, one is well-advised to select bees that are showing clear symptoms of
pesticide poisoning, (the quivering, the staggering walk) and such WHILE
STILL ALIVE.  This allows one to at least assure that one has a 100%
pesticide-poisoned sample.

But you aren't anywhere near done.  There is an inherent accuracy issue with
any Mass spec measurements, as there would be with any measurement.  The
technology gets better, the LoDs go down, and the accuracy of identification
of substances goes up, but I think that the USDA Gastonia lab or any other
lab that beekeepers might afford would describe their bee, honey, wax, and
pollen analysis as "screening" rather than "analytical-grade work", and the
amounts measured as "relative indicators of how many orders of magnitude the
pesticide dose was above the LoD" rather than absolute values to be taken at
face value.  (Mass spec, regardless of type, is very very very very good at
telling you what you have, and a lot less good at telling you exactly how
much of it you have.)

Further still, the processes used to determine an LD50 are not going to be
updated or supplanted by any process using data from pesticide kills or
pesticide screening of honey, wax, pollen, and bees, simply because so many
kills involve "massive overkill doses", and so many screenings are part of
the sales process for honey and other bee products, intended to prove that
only trace levels of this or that can be found.  As such, these data sets
are going to populate both far ends of the bell curve, and will rarely
include data that will allow one to dial in a mean, a median, or a centroid
on that bell curve.

But wait, there's more! LD50s are products of caged-bee tests, and don't
scale well out to a hive setting without a large list of assumptions that
will swamp out the data.  When one does an LD50, one knows exactly how much
poison is being given to each bee, and all one has to "measure" is the
status of the bee (is it alive or not).  With large numbers of bees in each
group being fed the same dose, and many fine graduations in doses, the
statistical work done on the data yields very reliable NOECs and LD50s.
Going the other way around, starting with a dead bee, and not knowing
everything to which it has been exposed, we run into several layers of
fuzziness and many opportunities to have small errors add up, subtract, and
do long division.

That said, I'd be happy to contribute what pesticide analysis data we have,
but sadly, our data consists of impressively long lists of "ND" (Not
Detected), as possession of as little as a bottle of Round-Up in NYC is
considered more gauche than wearing fur.   Yes, I'm gloating, you would too
if you could.




             ***********************************************
The BEE-L mailing list is powered by L-Soft's renowned
LISTSERV(R) list management software.  For more information, go to:
http://www.lsoft.com/LISTSERV-powered.html