>> But any "distribution" is a purely theoretical thing.

> We've performed nearly 10,000 washes of 300-bee samples in our operation, so speak from a practical standpoint.

And the average error rate of that sampling/counting process is what?
What's the confidence level?
What do the outliers look like?
More basic, when were 300-bee samples compared to anything more rigorous than the estimate gained from forked-open brood samples?

>  We find washes of 300 bees to be consistent and valuable.

A larger operation can afford to shrug at whatever losses they suffer due to the "small" inherent error rate.  That's the essential problem here.

> For Joe Smallscale, use my varroa model...

An inherently flawed binomial/Poisson distribution is simply never going to be up to the task of modeling a non-binary environment that has been over-simplified to a binary model.  Yes, it works well ON AVERAGE and OVERALL, but it cannot be trusted for small numbers of hives at the low levels of infestation where treatment is  required.   The model needs a clarification and disclaimer of the "your mileage may vary" type.

This is like the statistician who sees a fire in a wastebasket - "Pour water on it!" yells the chemist.  "Cover the basket and cut off the air!" Yells the physicist.  Yet the statistician lights 3 more wastebaskets afire.  "What ARE you doing?" yell the others.  "I need a larger sample size" says the statistician.

"Joe Smallscale" needs to get away from attempts at "sampling", and into SCREENING.  The move to "sampling" is fine for someone who needs to be only statistically assured of a tolerable colony loss level.  But "statistical assurance" means that one is certain of nothing.

             ***********************************************
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