Pre-Analysis Power Statement¶
Design contract — instantiates Hypothesis Test Power Calibration
Records the target effect, error budget, frame, and interpretation boundaries before data collection, turning power calibration into a pre-committed design contract.
The single idea that makes this mechanism itself and not any of its computational siblings: it computes nothing — it commits the design in writing, in advance. Every other mechanism in this family produces a number: a required N, a curve, a grid. This one produces a document that locks in, before any data exist, what the study is powered to detect and how its results will be read. It states the null, alternative, and estimand; the smallest effect worth detecting; the alpha and target power; the assumptions the power rationale rests on; and — its defining clause — the interpretation boundaries: what a significant result will and will not license, and what a non-significant result may be taken to mean given the design's sensitivity. Timestamped and filed before enrollment, it converts power calibration from an after-the-fact justification into a contract the analysis must honor. Its whole value is temporal: the same words written after the data are worthless.
Example¶
A clinical team is about to launch a randomized trial of a new inhaler for moderate asthma. Before the first patient is enrolled, they file a pre-analysis power statement in the trial registry. It names the estimand precisely — the difference in mean rescue-inhaler uses per week between arms, in adults with moderate persistent asthma, at twelve weeks — and fixes the minimum clinically important difference at a reduction of one use per day, the smallest change the pulmonologists agree would alter prescribing. It records alpha at 5% two-sided, target power at 90%, the variance assumption behind the sample size, and the primary analysis model. Its binding clause is the interpretation rule: the primary endpoint is this contrast at this timepoint; any subgroup or secondary finding is explicitly labeled exploratory in advance. Months later, when a post-hoc subgroup looks striking, the statement is what keeps the team honest — that comparison was never powered and was pre-declared exploratory, so it generates a hypothesis, not a claim.
How it works¶
- Write the frame first. State the null, alternative, estimand, population, endpoint, and timepoint precisely enough that no later reinterpretation can quietly substitute a different comparison.
- Fix the meaningful effect and the budget. Record the smallest effect worth detecting and the alpha and target power, so the sensitivity target is on record, not negotiable after the fact.
- Attach the assumptions. Note the variance, attrition, and analysis model the power rationale depends on — importing, not deriving, the numbers the calculation mechanisms produced.
- Pre-commit the reading. Declare what significant and non-significant results will license, and which comparisons are confirmatory versus exploratory — then timestamp and lodge the document beyond later editing.
Tuning parameters¶
- Specificity of the frame — how tightly the estimand and endpoint are pinned; tighter wording blocks bait-and-switch but leaves less room for legitimate mid-study adaptation.
- Confirmatory/exploratory split — how many analyses are designated confirmatory; more confirmatory endpoints demand multiplicity control and more power.
- Amendment protocol — whether and how the statement may be revised before unblinding, and how changes are logged; strict rules protect credibility but reduce flexibility.
- Registration venue — a public registry, a journal's registered-report track, or an internal record; more public venues raise the cost of quietly deviating.
- Interpretation-rule sharpness — how explicitly a null result is bounded by detectable magnitude; a sharper rule prevents "no effect" over-reading but requires stating the design's limits plainly.
When it helps, and when it misleads¶
Its strength is that it makes power calibration reproducible and enforceable: by fixing the meaningful effect, the error budget, and the reading in advance, it turns "the study had enough data" from a vague assurance into an auditable commitment, and it blocks the most common way power stories rot — quietly moving the goalposts after the results are in. It misleads if mistaken for rigor by itself: a statement can be precisely written and still enshrine an optimistic variance or a meaningless effect threshold, so a filed document is only as good as the calibration behind it. The misuse it most directly guards against is HARKing — presenting a hypothesis discovered in the data as though it had been predicted, the very move a timestamped frame forecloses.[n1] The guarding discipline is to treat the statement as the frame the analysis must live inside, to log any deviation openly, and never to let its existence substitute for honest power arithmetic.
How it implements the components¶
pre_registered_interpretation_rule— its defining clause: the pre-committed rule for what significant and non-significant results may be taken to mean, filed before data.null_alternative_and_estimand_frame— it fixes the exact contrast, population, endpoint, and timepoint on the record, protecting against later reframing.decision_relevant_effect_threshold— it names the smallest effect worth detecting in advance, so power is argued from a pre-specified meaningful effect, not the realized one.error_rate_budget— it records alpha and target power as binding design commitments rather than after-the-fact choices.
It builds no operating_characteristic_model and runs no sensitivity_scenario_grid: it imports the power arithmetic that Closed-Form Power Calculation, Simulation-Based Power Analysis, and Power Sensitivity Grid produce, and records — rather than estimates — the noise_and_variance_profile that Pilot Variance Estimation measures. It is the only sibling that computes nothing.
Related¶
- Instantiates: Hypothesis Test Power Calibration — the contract that pre-commits the calibration and its reading.
- Consumes: Closed-Form Power Calculation or Simulation-Based Power Analysis — supplies the power numbers the statement records.
- Sibling mechanisms: Closed-Form Power Calculation · Simulation-Based Power Analysis · Minimum Detectable Effect Table · Operating Characteristic Curve · Power Sensitivity Grid · Pilot Variance Estimation
Editorial Notes¶
Form Classification¶
Form family: Rule, Policy & Commitment
Rationale: The mechanism precommits the estimand, frame, meaningful effect, error budget, power target, assumptions, and interpretation boundary before data exist.
Nearest alternative: Representation, Specification & Plan — The contract is documented, but its operative force is the standing constraint preventing later reinterpretation.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Prospective declaration of target effect, error rates, and sample-design assumptions belongs to statistical power analysis and preregistration.
Review outcome: Independent reviewer agreement; high confidence.
Notes¶
[n1] HARKing — Hypothesizing After the Results are Known — is presenting a post-hoc finding as if it had been predicted in advance. A timestamped pre-analysis statement forecloses it structurally: the confirmatory hypotheses and their interpretation rule are on record before the data can suggest new ones. ↩