Preregistering secondary data: prior access and analysis decisions

What can still be prospective?

An existing dataset can support a prospectively recorded analysis, but the claim must match actual timing and knowledge. OSF offers a secondary-data template. Distinguish inspecting a dictionary, marginal distributions, relationships and the target analysis.

Disclose a previously inspected correlation. A later plan documents future decisions but cannot make completed exploration independently confirmatory. Where feasible, reserve genuinely unseen data or a later wave for evaluation.

Record prior access

List person, date, file version, variables and inspection performed. Published analyses and information conveyed by collaborators can influence choices too; file access alone differs from outcome inspection.

In the hypothetical case, the analyst saw column definitions and a public recruitment summary but not outcome values. Support that statement with a work log instead of a vague assertion of blinding.

Define the analytical sample

Specify dataset release, time window, unit, eligibility, repeated observations, weights and permitted use. Document derived variables, missingness and exclusions before examining the target relationship. Do not choose exclusions to favor a finding.

Respect complex survey design where applicable. Fixed data availability does not establish adequate precision or power; plan sensitivity and acknowledge estimation limits.

Write decisions that code can implement

For a teaching question about study hours and second-wave scores, specify variable names, units, justified adjustment, model, failure conditions and reporting. Predictive evaluation requires test separation before learned preprocessing and tuning.

Attach code version and expected inputs and outputs. Define a sensitivity model explicitly instead of keeping unrestricted post-result model choice.

Registration, amendments and reporting

Keep the dated registration. For each amendment record what changed, why, when and which results were known. Cite the record and separate principal, sensitivity and exploratory analyses.

An authorized colleague should reconstruct the decisions from protocol, dictionary and code. The teaching case is not an actual survey; permissions for real data remain separate.

Completed teaching worksheet

This is a hypothetical teaching case, not observed data, an actual review or a publication acceptance. Numbers illustrate decisions.

Deliverables and completion checks

• Complete prior-access record.
• Explicit sample, variable and model rules.
• Dated registration, amendments and rerunnable code.

Version and scope references: OSF Support — Registration templates · Center for Open Science — Preregistration