Study guide: Publish a Reproducible Data Argument
For 12 completed attempts out of 15 supplied attempts, write a reproducible result with provenance and a limit. Separately classify an unavailable phone test.
Ideas to explain in your own words
- Provenance records where data came from and what happened to it. A source, date, definition and transformation log let another learner interpret a result. A summary without those details may be hard to reproduce. The lesson uses invented datasets so no real participant information is disclosed.
- A percentage is a numerator divided by a specified denominator. Different denominators answer different questions. A completed-attempt rate is not the same as a unique-person success rate. Our supplied quantities represent attempts only, and no causal inference is established by calculating a rate.
- Reproducing a result means obtaining the stated output from the supplied data and method. It does not prove the method fits every population or that an observed change was caused by a platform. A report should distinguish calculation checks, data representativeness and causal evidence.
- A useful report can state a verified result and its practical limits together. Unverified is not the same as false, but it should not be marked passed. Describe the exact missing check and the evidence that does exist. Our report cases concern a fictional analysis release, not a claim about LifeStarter’s effectiveness.
Useful vocabulary
- provenance
- The data’s origin and processing history.
- denominator
- The full set used for a rate.
- reproducible
- Another person can obtain the result from the stated method.
Watch your reasoning
- A polished chart is not a substitute for provenance.
- Do not call an attempt rate a people rate.
- Correct arithmetic does not establish causation or population representativeness.
- Do not relabel an unavailable test as a successful test.
A worked review example
Fictional dataset supplied by this lesson. Which provenance element is described? The supported answer is Source. Record the origin, measurement definition and transformation separately. A polished chart is not a substitute for provenance. Provenance records where data came from and what happened to it. The supplied facts, rather than an assumed real-world detail, determine this paper case. Compare the recorded action or value with the stated rule before choosing an answer. Re-run supplied calculation matches expected output. Classify the report item. The supported answer is Verified in stated check. Separate verified observations, known failures, missing tests and explicitly deferred work. Do not relabel an unavailable test as a successful test. A useful report can state a verified result and its practical limits together.
Make a practice plan
- Make a four-row summary, one row for each lesson, with its computing idea and one worked decision.
- Explain one tempting mistake using the actual supplied example.
- Complete the new unit review before consulting its explanations.
- Try the transfer task on paper and compare your evidence against the model response.