NINETY-NINE topics are dealt with, each containing a rule of thumb, the basis for which is clearly discussed. Of course, it is not possible in the space provided to consider all of these. So I have done what any reasonable statistician would do: I have taken a nonrandom sample of ten as follows. Estimate the standard error with sample range/sample size. A table gives the divisors for the sample range to estimate the standard error of the sample mean. In the Basis of the Rule, a half-page discussion of its origin and its accuracy are given. Overlapping confidence intervals do not imply nonsignificance. Calculations are given to demonstrate that 95 percent confidence intervals can overlap as much as 29 percent when there is a difference in two means that is significant at the 0.05 level.
展开▼