Qualification: Lessons from the first two years of operating a study registry
Confirmatory analysis depends on when analytical discretion closes. Watt and Kennedy require every analysis decision capable of affecting confirmatory results to be made before data collection. They include statistical methods, evidence criteria, data transformations, and exclusion rules. This provides an external benchmark for Kaal's design. Kaal fixed directional hypotheses, a bootstrap estimator, exclusions, a randomization audit, a data-quality gate, and the confirmation rule before confirmatory computation. The benchmark also narrows the claim. Watt and Kennedy use a stronger timing rule than Kaal's statement. Their standard closes analytical discretion before data collection, while Kaal's source-bound claim establishes closure before confirmatory computation. The article does not audit Kaal's preregistration, determine when his data became available, reproduce the bootstrap, or validate the stated gates. It therefore supports the importance of advance specification and the listed categories of decision. It does not prove that Kaal met the article's stronger prospective standard. That limitation is substantive. A confirmatory label cannot substitute for a fixed analysis plan. Kaal's implementation and empirical results remain separate source-bound claims.
research-methodsscholarly-growth-coveragescholarly-literaturepreregistrationconfirmatory-researchanalysis-plan