Power of a Test Math Example 4

Follow the full solution, then compare it with the other examples linked below.

Example 4

hard
A study fails to reject H0H_0 and concludes 'there is no effect.' Critique this conclusion using the concept of power, and explain what information is needed before accepting this conclusion.

Solution

  1. 1
    Failure to reject H0H_0 ≠ evidence that H0H_0 is true; it only means insufficient evidence to reject
  2. 2
    Low power: if power was 30%, we'd miss real effects 70% of the time; failure to reject could be due to small n, not because no effect exists
  3. 3
    Information needed: (1) Was power ≥ 80% at a meaningful effect size? (2) What effect sizes would this study have detected? (3) How large is the confidence interval?
  4. 4
    Better conclusion: 'We found no significant evidence of an effect; however, with this sample size, we could only detect effects of size X or larger'

Answer

Failure to reject ≠ no effect exists. Need power analysis to determine if the study had adequate sensitivity.
'Absence of evidence is not evidence of absence' — a classic statistical principle. A study with low power frequently misses real effects. Reporting power, effect size estimates, and confidence intervals is essential for interpreting non-significant results appropriately.

About Power of a Test

The probability that a hypothesis test correctly rejects a false null hypothesis. Power =P(reject H0H0 is false)=1β= P(\text{reject } H_0 \mid H_0 \text{ is false}) = 1 - \beta, where β\beta is the probability of a Type II error.

Learn more about Power of a Test →

More Power of a Test Examples