econometrics · Hybrid cohort
Regression Diagnostics in Practice
Residual plots, influence points, and specification checks you can repeat before reporting estimates.
ZAR 890 · 6 weeks · async + live labs
Software: R · Method focus: regression · Level: intermediate · Assessment: labs + memo
Description
This course walks through heteroskedasticity-robust errors, leverage, and Cook's distance with annotated do-files and R scripts. You compare nested models with disciplined F-tests and document what changed when you add controls.
What is included
- Weekly diagnostic checklists aligned to published tables
- Annotated code for robust and clustered SE comparisons
- Office-hour review of one working paper section
- Template language for limitations in appendices
- Peer review of diagnostic plots with mentor notes
- Short readings on when to stop adding controls
Outcomes
- Produce a defensible diagnostics appendix for a linear model
- Explain leverage-driven changes without overfitting language
- Choose HC variants appropriate to your data structure
Dr. Naledi Mokoena
Former RA for a labour policy lab; teaches diagnostics-first workflows.
Reviews
The leverage walkthrough finally matched what our PI asked for in referee responses — especially the Cook distance paragraph in week 4.
Clear labs. I wanted one more week on nonlinearities, but the diagnostic checklist is now pinned above my desk.
Questions
We focus on linear-in-parameters models with transformations. Logit and count models live in a separate track.
We include missingness patterns and simple imputation sensitivity, but full multiple imputation is not included.
Yes, after passing two lab checkpoints and a short interpretation brief.