Skip to content

beyond-null vignette: add pre-specification note for the margin and a dose-composition caveat #25

Description

@ismayc

Two rigor gaps in the vignette, both aimed at the applied and clinical audience the package targets:

  1. Pre-specification. The margins used (the 3-unit bar in Showcase 1 and the 20% rule in Showcase 3) read as chosen for the examples. For a superiority or threshold test, the margin has to come from domain knowledge and be fixed before seeing the estimate, not selected afterward. Add one sentence making that explicit, since a margin chosen after the estimate is in view is exactly what this audience is trained to distrust.

  2. Dose composition of the dynamic threshold. In Showcase 3, overall_avg is computed from the full dataset across all three doses, so the threshold shifts with the experiment's dose composition, not only with sampling noise. A design weighted toward higher doses would move the benchmark. This is a separate concern from the estimation-uncertainty footnote already added (which covers treating a data-derived threshold as fixed), and deserves its own clause.

Location: vignettes/usage/beyond-null.Rmd, Showcases 1 and 3.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions