Preregistration

Lock in your hypotheses and analysis plan before you collect data to guard against p-hacking and HARKing.

Preregistration means documenting your research plan, your hypotheses, your design, and your analysis, and timestamping it in a public archive before you collect or look at your data. It is the strongest single thing you can do to make a confirmatory study credible.

Why Preregister

When the analysis plan is fixed in advance, you remove the researcher degrees of freedom that quietly inflate false positives:

p-hacking: trying many analyses and reporting only the one that crossed significance.
HARKing: presenting a post hoc hypothesis as an a priori one. See Research Design.
Selective reporting: dropping conditions, outcomes, or exclusions that did not work out.

A preregistration lets a reader distinguish what you predicted from what you discovered. Both are valuable, but only the first counts as confirmatory evidence.

What to Include

A useful preregistration is specific enough that someone else could run your analysis without asking you anything:

Section	What to specify
Hypotheses	The exact predictions, stated directionally where possible
Design	Conditions, manipulations, and how participants are assigned
Sample	Target size, how you arrived at it, and stopping rule
Measures	Every variable and exactly how it is computed
Exclusions	Rules for removing participants or responses, written before you see the data
Analysis	The specific test for each hypothesis, including how you handle covariates

Decide your participant exclusion rules in advance. Removing data after seeing results, even for defensible reasons, reopens the door to bias.

Where to Preregister

Independent registries timestamp your plan so it cannot be edited silently:

OSF (Open Science Framework): flexible, widely used, supports embargoes.
AsPredicted: a short standardized form, good for simple designs.

Registered Reports

A Registered Report goes one step further. You submit your introduction, methods, and analysis plan to a journal for peer review before collecting data. If the plan is sound, the journal grants in-principle acceptance, committing to publish the results regardless of how they turn out. This removes publication bias toward positive findings and rewards good design over lucky results.