Data Analysis Dataset Safety Policy
Applies to: all datasets, notebooks, downloads, examples, charts, and reports in this course
Rules
- Prefer fictional or synthetic data created for this course.
- Do not use real personal, confidential, employer, client, patient, financial, authentication, legal-case, or security-sensitive data.
- Do not ask learners to upload or paste private data.
- Do not download live remote data during learner exercises.
- Record creator, date, version, licence, permitted use, and checksum for each dataset.
- Include a data dictionary for each dataset.
- Include expected schema, missing-value rules, and known limitations.
- Keep raw/source data read-only.
- Save cleaned outputs under new names.
- Remove a dataset from publication if rights, provenance, privacy, or safety cannot be demonstrated.
Synthetic data standard
Synthetic data must feel realistic enough to teach the concept but must not be traceable to real people or companies. Fictional identifiers must not resemble real emails, phone numbers, addresses, account numbers, payment data, or secrets.
Allowed imperfections
Synthetic datasets may include:
- missing values;
- duplicate rows;
- inconsistent category labels;
- invalid numeric ranges;
- date parsing problems;
- delayed records;
- outliers;
- join-key mismatches.
These imperfections should be intentional and documented so they teach a lesson.
Review before publication
Before publication, each dataset must pass:
- schema check;
- missing-value check;
- duplicate check;
- private-data scan;
- secret/token scan;
- provenance review;
- notebook rerun; and
- chart/output review.