Hands-On
Everything so far has been preparation for this: a real dataset, a real question, and a real problem that no method can fix.
About this section
You will work a genuine study end-to-end — from haplogroup calls to a hypothesis test, and then to the discovery that the test cannot be interpreted.
This is not a contrived teaching example. It is a real cohort from Botswana, and the flaw you will find is a real flaw in a real analysis.
The session
| Activity | Time | |
|---|---|---|
| 1 | Case study: mtDNA haplogroups & HIV status | 90 min |
| 2 | Group discussion: what would you do next? | 30 min |
| 3 | Bring your own data clinic | 60 min |
What you will do
Load 338 participants with HaploGrep 3 haplogroup calls and HIV status. Wrangle, visualise, build a contingency table, run the test — the standard workflow from Section 2.
Then run one more line of code and discover that the study design is confounded beyond repair.
Every teaching dataset you have ever met was chosen because it behaves. That is why analysts are unprepared for real data.
The skill worth having is not running a test. It is knowing whether the test means anything. You cannot learn that on clean data.
Bring your own data
The last hour is unstructured. Bring a dataset and a question. The most useful thing we can do is help you find the problem in your data before a reviewer does.
The dataset you are unsure about is more valuable here than the one you have already published.