preregistering-analysis

To prevent p-hacking and HARKing, you must preregister your hypotheses, analysis plans, and decision rules before observing outcomes. This Skill enforces a LOCK, FREEZE, EXECUTE, SEPARATE protocol to freeze predictions with a timestamp before data analysis begins.

A preregistration protocol for confirmatory analysis requires writing hypotheses, primary analysis, predictions, and stopping rules before looking at results. It commits a timestamped preregistration to your repository path and clearly labels any unregistered follow-ups as exploratory.

You create a timestamped preregistration for an observational study by documenting your hypothesis testing plan and decision rules, then committing them to a repository path like docs/science-superpowers/preregistrations/. This freezes your analysis plan before executing the study.

Yes, you can use this preregistration workflow for any confirmatory-analysis project, including experiments and observational studies. It applies across hypothesis testing, p-values, and reporting claims by enforcing explicit documentation of predictions before results.

The best way to separate exploratory analysis from registered predictions is the EXECUTE and SEPARATE protocol. You run exactly the registered analysis first, then clearly label any unregistered follow-ups as Exploratory in your final reporting claims.

Yes, you need to define stopping rules and a multiplicity plan as part of the LOCK phase before preregistering your analysis plan. Writing these decision rules before observing outcomes is required to prevent post-hoc hypotheses and ensure reproducibility.