AtlasSession 2L6: Hypothesis Testing and Statistical Testsp.25–33
Tests, p-values, significance
- 3 earlier topics
- Comparing groups: real difference or chance?
- What is a hypothesis
- Test, null and alternate hypotheses
- Alpha error
- Tests, p-values, significance
- Next: p-value misconceptions
Fig. 1
Difference in mean BV/TV at 8 weeks, strontium-doped minus undoped bioactive glass, rat condyle defects (percentage points)
Fig. 1. One hump: the results you would get if doping made no difference. The gap you saw, 6.00 pp, is marked. Results as far out, or further turn up about 3 in 100 times when nothing is going on.
- This far out, if nothing is going on
- about 3 in 100
- Rarer than 5 in 100?
- yes
Two-Curve Test Visualiser · First look
Level 1 · First look · Lecture 6 · p.25–33
Ask 'if nothing were going on, how surprising is this result?' and read the answer as a p-value.
Explain a p-value as the probability, under the test hypothesis, of a difference at least as large as the one observed, and call a result significant against a pre-set threshold.
Surprise meter (the step you are on)
A friend's coin gives 9 heads in 10; if the coin were fair, how often would that happen?
A friend’s coin lands heads 9 times in 10. If the coin were fair, 9 or more heads would turn up only about 1 time in 100, so you would be surprised.