Good practice for planning a study

Defaults for the choices a study makes before it is run: which measure, which test, which design fits a kind of study, and why. Each default cites the papers it rests on. They are proposals, never requirements: a study may choose otherwise, and says why. An assistant planning a study with you proposes them and asks you about the rest (plan a study with any assistant).

What a guide holds

  • A principle first: the reason the defaults follow from, stated so that it applies beyond the cases listed.
  • Entries, each under a stable anchor that plans cite (study-planning/prediction#mean-forecasts):
    • When: the situation it covers.
    • Default: what to use.
    • Why: the argument in one or two sentences.
    • Avoid: common choices that do not fit, and why.
    • References: papers by DOI, and by claim where the paper is on EntryLib (entrylib:2026.000001@v2#C4).
  • Only choices with a methodological argument behind them. A setting that does not change conclusions, or a convention without a reason, is not a default.
  • A precedent is not a reason. That a competition or a well-known paper used a measure says what it did, not why it fits another study.

Changing a guide

  • By pull request, with the references that support the change. Disagreement in a field happens here, in the open.
  • Each guide has a version; a changed default is a new version, listed at the end of the guide. A plan cites the entry, and the version it used is the one current on the plan's date.
  • Anchors never change meaning. A default that is withdrawn keeps its anchor, marked withdrawn, with the reason.
  • New guides are welcome from people who work in the field they cover.

Licence: CC BY 4.0 (Creative Commons Attribution 4.0 International, see LICENSE and https://creativecommons.org/licenses/by/4.0/).

From entrylib/study-planning at 104ba1c (2026-11-13), as the installed claimset package carries it.