Week 12 · Suggested 90 minutes · First-pass draft
Confounding and regression in the same room
Use a historical Simpson's-paradox example and a synthetic regression to separate descriptive patterns from causal claims.
Applied assignment, not a stored grade. Instructions and model checks are provided for self-study or instructor adaptation. Work is not transmitted to this site.
What you will do
- Compute pooled and size-stratified historical success proportions for the two kidney-stone procedures.
- Explain the composition change responsible for the reversal without asserting treatment superiority.
- Fit a straight line to the supplied synthetic curved data and examine the residuals as curvature increases.
- Explain what a fitted slope and R-squared do not establish about causation or prediction.
What to submit
Two interpretive figures/tables and a research-design critique.
Self-check before finishing
Open surgery has a higher observed rate within each size group, whereas the other procedure has a higher pooled observed rate.
Assigned reading
Investigate Simpson's paradox, fitted lines, residuals, and slope uncertainty.
- Lesson 35: simpsons paradox
- Lesson 36: least squares regression
- Lesson 37: residual diagnostics
- Lesson 38: regression coefficient inference
Instructor discussion guidance
Reject clinical recommendations inferred from historical observational counts; inspect the severity composition.
Print a predict–experiment–explain worksheet → · Open browser R →