Preregistration

Preregistration is the practice of publicly documenting a study's hypotheses, planned sample size, and analysis approach before any data is collected or examined, typically time-stamped in a public registry.

A preregistration commits the researcher to a specific analysis plan in advance, so that once results are published, readers can distinguish confirmatory findings that tested a stated hypothesis from exploratory findings discovered by looking at the data first; any deviation from the original plan is expected to be disclosed and justified rather than quietly folded in as if it had been planned all along.

Preregistration directly addresses practices like "p-hacking" and "HARKing" (Hypothesizing After the Results are Known), where a researcher tries many analyses and reports only the ones that reach statistical significance, which inflates the rate of false positives even when each individual hypothesis testing step uses a standard p-value threshold correctly. Pitfalls include that preregistration constrains but does not guarantee good research, a vague or overly flexible plan still leaves room for selective analysis, and exploratory analysis is not illegitimate in itself, it simply needs to be reported as exploratory rather than presented as if it had been predicted in advance.