Module 01 Summary

What this module established

A statistical project needs one folder a data project does not: the analysis plan, dated before the analysis. Choices made after seeing the data — which outcome, which test, one-sided or two — are unfalsifiable, because a version of them always exists that gives the answer you hoped for.

Carry forward

• Scripts compute and save; reports explain and cite. The report cannot then contain a number no code produced.
• Seed once at the top. An unseeded bootstrap gives a different interval every run, and a seed inside a loop gives no variation at all.
• broom turns test objects into data frames, so twenty tests become one table and no number is ever retyped.
• Every exclusion is a rule with a reason and a printed cost. Done by hand it is invisible, and it moves the headline figure.

Before moving on

Move on when your plan is dated, your seed is recorded, and your results table contains every comparison you ran.

Mark complete