GurjasEvidence & Policy Analytics
Research Library · Research integrity

Preregistration does not replace an analysis plan

A timestamped registration can improve transparency only to the extent that the registered decisions are specific enough to distinguish confirmatory from later exploratory choices.

Direct answer

Preregistration records planned decisions before a defined stage, but it does not guarantee that the analysis plan is complete, appropriate or followed. Review the timing, specificity, amendments and final reporting together.

What to inspect

  1. Timing. Identify whether registration occurred before data collection, before access to outcomes or after some analysis.
  2. Specificity. Look for operational hypotheses, outcomes, exclusions, transformations, models and decision rules.
  3. Version history. Preserve amendments with dates and reasons rather than overwriting the original plan.
  4. Final alignment. Compare the publication with the registered plan and label justified deviations.

Registered reports are different

A registered report is a journal format in which the protocol is peer reviewed before results are known and may receive in-principle acceptance. A general preregistration platform records a plan but does not by itself provide that editorial review or commitment.

Worked use case

A registration says only that “regression analysis” will be used. The paper later reports several models and subgroup analyses. The registration supports timing, but not a claim that every modelling choice was prespecified. Report the gap and classify the additional analyses transparently.

Evidence boundary

Preregistration is not a universal quality certificate. Poorly specified plans can still be timestamped, and legitimate deviations can improve a study when they are explained and separated from confirmatory claims.

Primary sources

  1. Center for Open Science. Preregistration guidance. Retrieved 5 August 2026.
  2. Center for Open Science. Registered Reports. Retrieved 5 August 2026.