ABSTRACT
Both the Neyman–Pearson and the Fisherian views of statistics recognize three major divisions – estimation, hypothesis testing, and confidence intervals. These divisions represent a fragmented, piecemeal approach to analyzing and interpreting statistical evidence. Each of the three in turn has its own complicated set of concepts and techniques that must be used expertly. For example, what it means when a hypothesis test does not call for rejection must be understood in terms of: size and power; the distinction between ‘not rejecting’ a hypothesis and ‘accepting’ it; the distinction between ‘statistical significance’ and practical (or scientific) significance; the way that sample size affects the meaning of p-values, etc. We have seen that such matters are complicated enough that experts cannot agree.
