74 CHAPTER 6 Hypothesis Testing
Figure 6.1. Power functions for a normal distribution with mean δ and sample sizes 25
and 100. Null hypothesis δ = 0.
Power Function for two sample sizes
Power
Alternative Mean
–1.5 –1 –0.5 0 0.5 1 1.5
- 2
1
0.8
0.6
0
0.4
0.2
N=100 N=25
6.2 ONE - TAILED AND TWO - TAILED TESTS
The test described with the power function in Figure 6.1 is an example
of a two - tailed test. Two - tailed tests are test where we consider both
δ > 0 and δ < 0 as part of the alternative. A one - tailed test is a test
where only one side is of interest for the alternative. So, for example,
if you want to show drug A is better than drug B at lowering cholesterol,
we would only be interested to see if drug A had a larger drop from
baseline in cholesterol than drug B. Then, if we take δ = change from
baseline for A − change from baseline for B, we are interested if δ < 0.
But δ > 0 is no more interesting than δ = 0. So in this case, δ > 0 is as
much a part of the null hypothesis as δ = 0. There are also cases where
δ ≤ 0 is not interesting, and is included in the null hypothesis because
we are only interested if we believe δ > 0.
6.3 P - VALUES
The p - value is simply the probability of getting a value as extreme or
more extreme than the actual value of the observed statistic when the