Suppose estimates of all pairwise comparisons are of primary importance. What would be the required sample sizes if the precision of all pairwise comparisons is to be + – 3.0, using the Turkey procedure with a 95 percent family confidence coefficient