Week 4 · Cause and effect in the RCT. Hypothesis testing. t-tests
Null and alternative hypotheses, the two error types and the p-value, applied to two field experiments
The week turns the causal language of Week 3 into a testable procedure. It introduces hypothesis testing in the Neyman-Pearson framework and then works through two randomised experiments, a school deworming programme in Kenya and a conditional cash transfer, to see how checks-and-balances tables and t-tests are used to read an RCT.
Slides
The 2024/25 deck is titled Cause and Effect in the RCT. Hypothesis testing. T-tests. Examples. Central Limit Theorem. The 2023/24 PDF was shared as Week 04. Cause and Effect in the RCT. Hypothesis testing. T-tests and is longer (78 pages against 60); the Central Limit Theorem is not in its title.
| Material | WS 2024/25 | WS 2023/24 |
|---|---|---|
| 🖥 Interactive slides | Week 04 | not rendered (sources were revised for 2024/25) |
| 📄 PDF from Ilias | Week 04 (version 1).pdf · 60 pages · added 5 Nov 2024 Week 04 (version 2).pdf · 60 pages · added 27 Nov 2024 |
Week 04. Cause and Effect in the RCT. Hypothesis testing. T-tests.pdf · 78 pages · added 10 Jan 2024 |
| 💻 Source | mk68-04-slides.qmd, which includes hypothesis-testing.qmd, deworming.qmd and cct-2-simple.qmd |
mk68-04-slides.qmd |
What the deck covers
- Hypothesis testing: the Neyman-Pearson approach, the steps of a test, the two errors a test can make, the p-value and the level of significance; examples on stunting and education and on raven colour; homework.
- Examples of hypothesis testing from a nutrition education RCT: research design, checks and balances, balance of household size, and how to identify whether the intervention helped.
- Effects of a deworming pill: the Primary School Deworming Project timeline, checks and balances between groups, immediate effects on health and school participation, long-term economic effects, and costs and benefits.
- t-tests with conditional cash transfers: cause and effect, what makes the difference in enrolment rates a causal effect, good controls in an RCT, checking that ceteris paribus holds, the checks-and-balances table, the t-test algorithm with worked examples, and conclusions.
- Homework, readings (Chapters 1 and 2 of Mastering ’Metrics) and the key concepts checklist: causal meaning of averages, random sample variation, type I and type II errors, significance level, p-value, and the standard error of a sample mean.
Data used
The examples in this week are read from the published tables of the studies discussed; the images are in materials/img.
Where the course goes next
← Week 3 · Bi-variate data. Correlation and causation · → Week 5 · Central Limit Theorem. t-distribution. Confidence intervals