Lectures

From the null world to the p-value, errors, power and intervals: what a test is, what it promises and what it does not. Chapters 1 to 5.

1. The null world 12 min
2. Hypotheses are about parameters 8 min
3. Switching tails after peeking 8 min
4. The shaded tail 10 min
5. What a p-value is not 8 min
6. Two worlds: errors and power 10 min
7. Run the experiment 1,000 times 8 min
8. One calculation, two views 10 min
9. The one-sample t explorer 8 min
10. Exercises 8 min

The everyday tests on Helion's and Meridian's data: t for one mean, Welch for two, the paired test, proportions, chi-square and variances. Chapters 6 to 12.

1. z against t 8 min
2. The twelve steps 12 min
3. Two samples pulling apart 10 min
4. Pooled against Welch 8 min
5. Pairs collapse into differences 10 min
6. One proportion: bars and the curve 8 min
7. Two proportions and the pilot 10 min
8. Observed against expected 10 min
9. Where χ² and F come from 8 min
10. Exercises 6 min

ANOVA and Tukey, rank and permutation tests, choosing a test, then the part most courses skip: effect sizes, power planned in advance, and multiple testing. Chapters 13 to 18.

1. Between against within 10 min
2. Which regions? Tukey's intervals 8 min
3. Shuffle the labels 10 min
4. The chooser's path 10 min
5. Significant against important 8 min
6. The underpowered study 8 min
7. The power curve: plan the pilot properly 10 min
8. The garden of forking paths 10 min
9. Peeking 8 min
10. Exercises 8 min

The two capstones worked in small groups: Helion's A/B readout, gates and memo included, and Meridian's risk review with rates, paired models and equivalence. Chapters 19 and 20.

1. The experiment pipeline 10 min
2. The sample ratio mismatch check 10 min
3. H1: the interval, not the point 10 min
4. Break the randomisation 10 min
5. Exposure matters 10 min
6. Equivalence, and the estimand 10 min