Allen,

Great questions.  Here are some rough thoughts in return.

> Can you state the results numerically and also calculate the degree of confidence?

You can state the results numerically.  You have a 0% infestation of varroa and tracheal mites on adult bees within your operation.  The confidence you can place in this estimate, however, is more difficult. To know with what confidence you can make your prediction you must establish three things experimentally:
 1) The confidence that a 5 bee sample, taken from each colony, predicts the true infestation within the yard (e.g. compare your 5 bee sample to a 42d Apistan sticky board drop count).
 2) The confidence that restricting sampling to 1/6th of your apiaries predicts the true infestation throughout your operation (e.g. sample all the colonies and determine what loss in confidence would result from sampling only 1/6th of the yards).
3) The confidence, in your area, that if you treat or do not treat, you will experience economic loss (either in wasted treatments or in lost honey production or colony loss).

By gathering this data you can use standard statistical analyses to get an estimate of confidence.  What is more, with such data you can often model the consequences of taking more or less samples, either within colonies or between yards, and figure out which gives the best prediction for a given effort.

> Were sufficient bees examined to give reasonable certainty?

I seem to remember Rob Currie and Paul Gatien's work with varroa suggested that a sample of 300 bees from 5 colonies within a yard was able to predict a threshold for withholding treatment without incurring reduced honey or colony survival under Manitoba conditions.  I have never seen anyone test the sampling scheme you are proposing... it may have merit but would require extensive validation before it could be trusted (ie experiments which would establish the consequence of treating or not treating a yard based on the results of your proposed sampling scheme).

> What are potential sources of bias or error?  Do you suspect and errors in sampling?

If you really truly randomly select a sample, there is no bias, however there can be error.  Error is the random 'noise' the clouds the 'signal' of a prediction.  What do I mean by noise?  I mean unpredictable ups and downs between your sample result and the true infestation level.  Bias, by contrast, is not unpredictable, it is a consistently wrong estimate (e.g. always overestimating infestation by 2%).

With such as small sample there are many sources of unpredictable ERROR; you will never get bees from a colony that TOTALLY mirror the infestation of the whole colony (this error decreases as you sample more bees... until you sample ALL the bees in a colony, but then, alas, the colony is dead); you will never select the apiaries that exactly mirror the infestation of over apiaries.... nonetheless, as long as the error is relatively small, then your prediction will carry some weight.

As for BIAS, you are not collecting bees randomly from a colony.  At this time of the year you are getting them off the lid.  These bees may be CONSISTENTLY more or less infested than the bees in the rest of the colony and may CONSISTENTLY over or underestimate infestation... no problem as long as you know the magnitude of the consistency and take it into account when making your estimates.

> How much should we trust the results?

There is something to be said about getting so many consecutive zeros, but to be honest I would not want to go out on too much of a limb without data to validate your sampling method.

> Compare the cost of an IPM approach like this to routine blanket treatment without sampling.

This is the kicker.  Obviously, if your sampling method was true, and you had hardly any mites, then you could save $10,000 on unnecessary acaricide treatments and help delay the onset of resistance by potentially giving your mites a break from selection for a generation.  Did it take a 40 person hours to do the sampling?  Lets say it cost you $500.  In the best case you come out on top $9,500.  If the sampling was downright unpredictive, and you experience a 50% reduction in production and loose 20% of your colonies before you figure out what is happening... that would mean the new Winnabego is that much further an element of your dream life ;)

No but seriously, I just finished the Canadian Beekeeper Pest Management survey and less than 8% of respondents answered they 'monitor for varroa and treat when necessary'.  Everyone else is treating on a calender.  Monitoring has clearly been slow in being adopted as a management tool by beekeepers and it is good to be asking, 'why is this so'?

Adony



__________________________________________________
D O T E A S Y - "Join the web hosting revolution!"
             http://www.doteasy.com