> An IPM question:
With a degree in Mathematics/Computer Science I can confidently say that I
have forgotten more stat theory than most people ever learn!  The thing I
most remember about stat theory are there are 4 kinds of lies: white lies,
damned lies, maps and statistics!

So, your example:  2400 hives divided by 66 yards divided by 10 yards at
random to make up your sampling set, less the 4.5% dead hives gives
3.4727272727272727 (sorry Robt) hives from which you took 5 random bees
(which given a spring population in an average Alberta hive must be about
5.555E-4 of the total (give or take)), meaning your sample set is for all
intense and purposes, phi (the empty set).
> * Can you state the results numerically
Hoboken, New Jersey.

> and also calculate the degree of confidence?
E=0 with a standard deviation of 1.4

> * Were sufficient bees examined to give reasonable certainty?
No.  Actually, depending on where from within the hive you took the bees
(center of cluster? periphery of brood? landing board?) the answer is still
No.

> * What are potential sources of bias or error?
Where did you take the 5 sample bees per hive?  How randow was the yard
selection?  Were the locations flat and exposed?  Were they in low frost
pockets?  Totally random?  Who took the samples?  Were the samples taken by
the same beekeeper or different hands?  The answer is still "No" and
Hoboken.

> Do you suspect and errors in sampling?
I assume that's "any", in which case, "Yes" and Newark.

> * How much should we trust the results?
No at all.

> What, if anything, can we conclude from the tests performed here?
NADA!

> * Is any treatment indicated?  If so for what disease or pest?
No and alzheimers.

> * What further action, if any, is indicated?
Take sufficient samples from sufficient hives to have a more realistic
sample set.  Drive to Hoboken and start over.

> * The balance of the original samples are being preserved.
> At what point may they be discarded?
As soon as you finish reading this (assuming you've gotten this far).

> * Compare the cost of an IPM approach like this to routine
> blanket treatment without sampling.
Well, labor and transportation to get the samples and lab costs is still
probably significantly less expensive that blanket treatments.  Just be
ready (and I'm sure you will) to resample and act accordingly if any
conclusions you draw from this experiment (if you DO draw any conclusions
from this experiment) prove to be incorrect.

> Have fun.
Sure did.

Aaron Morris - thinking Hackensack (it's just a stone throw from Hoboken)!