Prespecification

Prespecification is the practice of defining the plan for a study before data collection or inspection of results. It involves clearly specifying the research question or hypotheses, the primary and secondary outcomes, the study design, population and sampling criteria, data collection methods, variable definitions, and the statistical analysis plan, including the primary analysis, model form, handling of missing data, and criteria for data inclusion or exclusion. The aim is to prevent bias arising from data-driven decisions after seeing the data, such as selecting favorable endpoints or analyses, a concern often associated with HARKing (hypothesizing after results are known) and p-hacking.

In practice, prespecification is typically documented in a protocol or preregistration record. Researchers may register plans on public platforms (for example ClinicalTrials.gov, OSF Preregistration) or publish a protocol before data collection. Changes to prespecified plans are allowed in many settings but should be clearly documented and justified through amendments or the reporting of exploratory analyses separately.

Applications and scope: Prespecification is common in clinical trials, epidemiology, and other areas where bias from flexible analysis is a concern. It supports reproducibility and interpretability, helps distinguish confirmatory analyses from exploratory analyses, and can facilitate peer review and regulatory assessment.

Limitations: Prespecification may constrain necessary methodological adaptation to unforeseen issues during a study, and overly rigid plans can hinder innovation. Therefore, many guidelines emphasize transparent reporting of any deviations, robust sensitivity analyses, and clear labeling of confirmatory versus exploratory findings.

See also: preregistration, research reproducibility, HARKing.