In a message dated 4/7/2010 9:17:37 A.M. Mountain Daylight Time,  
[log in to unmask] writes:

In  general, any test should involve at least 12 normalized colonies in  
each
group...
... the two most important metrics are  bee
population (frame strength, which would include mortality), and weight  
gain.





I agree with the 12 colonies as a goal - my stats say that 6 is probably  
the minimum number for most things- but if you have several variables, you  
may have to settle for less than 12 per subset due to time/cost  factors.  
However, as you know, we've seen pesticide studies where  subsamples of the 
same colony were considered to be replications (which they  clearly are not).
 
As per metrics, its a lot of work, but if there's any chance of a  
reproductive effect, you need to follow amount of brood and its survival.
 
Finally, I've given up trying to normalize colonies - they just revert  
back to strong, medium, weak, regardless of what you do - I've even gone as far 
 as using sister queens hand inseminated to the same drones, with no real  
improvement.   Lay out your colony acceptance criteria before  starting the 
trial, then rank all of the colonies, apportion them in equal  strengths, 
etc. to each treatment.  In other words, if you have 3  treatments plus a 
control, and if you have 48 colonies, when ranked from  strongest to weakest, 
you'd take the top four and randomly assign one to each  treat/control, then 
the next four, etc.
 
Jerry

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

Access BEE-L directly at:
http://community.lsoft.com/scripts/wa-LSOFTDONATIONS.exe?A0=BEE-L