Milestone map
Milestone map
2 milestones
Do the exercises and show your work
7–14 weeks (10 hrs/week)
Take Econometrics for free and do the problem sets (regression, inference, omitted variables, heteroskedasticity, instruments, time series), done in code. Keep your answers in one public place (a repo, a shared folder or a page) as you go, with your reasoning shown.
Proof required
Link your public folder with the problem sets (regression, inference, omitted variables, heteroskedasticity, instruments, time series), done in code, and show your self-check (14.32: problem sets and exams).
What gets checked
- Each estimate is reproduced in code the verifier can run.
- Interpretations say what the coefficient does and does not mean.
- Where assumptions fail, the failure is shown.
Common mistakes
- Answers copied from a key: verifiers read the reasoning.
- Numbers with no working.
- Asking for review before the self-check.
Powstik Guide
A public folder with your worked answers to the problem sets (regression, inference, omitted variables, heteroskedasticity, instruments, time series), done in code.
Steps
- Make a public folder or repo for this course.
- Do one chapter or problem set at a time and save it.
- Self-check: Do each problem set and exam under time, then mark yourself against the solutions where the course posts them. Where it does not, rework the numbers in a spreadsheet and check that every total ties out.
- Fix what the self-check shows, and say what you fixed.
- Ask three people who know the topic to check it.
Template
README Course: Econometrics (MIT OpenCourseWare) What is here: Self-check result: The hardest question and why:
What gets sent back
- Answers with no working.
- Self-check not shown.
Resources
Foundationstart here
Depthgo deeper
Masteryfor the dedicated
What a verifier looks for
- Each estimate is reproduced in code the verifier can run.
- Interpretations say what the coefficient does and does not mean.
- Where assumptions fail, the failure is shown.
- Pick two answers and check every step.
- Ask about one step; the answer should be clear.
You'll sign in first, then come straight back here.
Estimate one effect with public data
1–3 weeks (5–8 hrs/week)
Use what you learned. Make a notebook that estimates one relationship from a public dataset (for example schooling and wages, or rainfall and yields), with the model, the standard errors, a robustness check and a plain statement of what it cannot show. Put it where verifiers can open it. Use public information, or data you have permission to use; never publish a real person's private details.
Proof required
Link a notebook that estimates one relationship from a public dataset (for example schooling and wages, or rainfall and yields), with the model, the standard errors, a robustness check and a plain statement of what it cannot show, and a short note on what you decided or changed because of it.
What gets checked
- The data source is named and the code runs.
- Standard errors and the main assumption are handled correctly.
- The conclusion does not claim cause without a reason.
Common mistakes
- Invented or unsourced figures.
- A conclusion the evidence does not support.
Powstik Guide
A notebook that estimates one relationship from a public dataset (for example schooling and wages, or rainfall and yields), with the model, the standard errors, a robustness check and a plain statement of what it cannot show.
Steps
- Write one line: what you will show.
- Gather the real data, people or documents.
- Draft it, then read it cold the next day.
- Ask someone to mark the first claim they doubt.
- Fix it and ask three people to check it.
Template
Title: What this shows: Sources: Method: Result: Limits:
What gets sent back
- Unsourced numbers.
- Not clearly the person's own work.
- Claims more than it shows.
Resources
Foundationstart here
Depthgo deeper
Masteryfor the dedicated
What a verifier looks for
- The data source is named and the code runs.
- Standard errors and the main assumption are handled correctly.
- The conclusion does not claim cause without a reason.
- Check that the work is the person's own and the sources are real.
- Ask how a number was found.
You'll sign in first, then come straight back here.