>1. how many drone makes a sample that will be accurate for my purpose?
I would just do a forkful (I would use an uncapping fork) in three or four locations. The results in one place may not be the same as another. But also keep in mind they will have to replace the drones you remove and if you hadn't removed them they would have spent those resources raising workers, so it costs you the same as if they were workers...
>2. In that sample size how many mites are an acceptable before treatment is needed?
I don't treat at all. Ever. But if I were looking for an economic threshold, I would say if you see an occasional drone with one mite I wouldn't worry about it. If you see a lot of drones with multiple mites you probably are over the threshold.
When I was counting mites, I would do three things. Mostly because none of them are 100% reliable. They are affected by temperatures and other factors that are hard to pinpoint. Drones can be heavily infested in one location in the hive and not in another. So if you uncap some drones, do a sugar shake and do a natural drop count, you have some redundancy built in. Also, I think the TREND is more important than the number. If the numbers are rapidly growing you have more of an issue than if they are staying steady, no matter what those numbers might be.