Module 01 Activity

Scenario

You have been asked whether a change worked. Set the project up so the answer is checkable by somebody who does not trust you.

What you build

A statistical project with a dated analysis plan, importable analysis modules and a report that only cites.

Steps

1. Write the analysis plan first: question, outcome, unit, comparison, primary test, alpha and sidedness. Date it.
2. Create the folders, including one for the plan, and put the raw data where no code writes.
3. Put rng = np.random.default_rng(seed) once at the top and pass the generator down. Record the seed in the report.
4. Build the analysis table explicitly: the planned columns only, the grain set, grouping variables as Categorical with deliberate category order.
5. Encode every exclusion as a named constant with its reason, and print how many rows it removed from each group.
6. Tidy every planned test into one DataFrame with the estimate, interval and p-value, then write the interpretation separately.

Evidence to hand in

• The dated analysis plan.
• The folder listing.
• The analysis table, with what was dropped from raw.
• The exclusion rules and their per-group row counts.
• The results DataFrame covering every planned comparison.
• The interpretation cell.

Review checklist

• The plan predates the analysis and names the primary test and alpha.
• A single default_rng seeds the work and its value is in the report.
• Every column in the analysis table was named in the plan.
• Every exclusion is a named constant with a printed row count.
• The results table contains all planned comparisons, not only the significant ones.

Mark complete