AtlasSession 2L6: Hypothesis Testing and Statistical Testsp.9–17, 30
Test, null and alternate hypotheses
- 1 earlier topics
- Averages and spread, by eye
- Samples and populations
- Comparing groups: real difference or chance?
- What is a hypothesis
- Test, null and alternate hypotheses
- Next: Alpha error
Fig. 1
Hypothesis builder: decellularised dermal matrix vs standard care, complete healing of diabetic foot ulcers at 12 weeks
A filled dot belongs to its side: a difference of exactly 10 is in the other possibility.
A trial starts by assuming that nothing special is going on, like ‘innocent until proven guilty’, and asks whether the evidence overturns that.
Even if the matrix makes no difference, two groups of ulcers given the same care still heal at slightly different rates, by chance alone.
Clinicians only care about a gain of 10 percentage points or more in healed ulcers.
Starting assumption: the matrix improves healing by less than 10 percentage points.
The other possibility: it improves healing by 10 percentage points or more.
- Smallest difference that matters
- 10 percentage points
Explorable Prose · First look
Level 1 · First look · Lecture 6 · p.9–17, 30
Say what the 'nothing special is going on' starting assumption is, and why two groups differ a little even then.
Write the test (null) and alternate hypotheses, including the course's MCID-framed form, and explain why sample means differ even when the null is true.
Innocent until proven guilty (the step you are on)
A trial starts by assuming 'no real difference' and asks whether the evidence overturns it.
Like ‘innocent until proven guilty’, the trial starts by assuming the matrix makes no real difference, a gain of less than 10 points, and asks whether the evidence overturns that.