🔬 Research
OSF Preregistration Writer: Hypotheses, Sampling Plan, Exclusion Rules, and Confirmatory Analysis Plan Before Data Collection
Turn a lab's study plan into a complete OSF Preregistration: testable hypotheses with directions, design and randomization, sample size with its power rationale and stopping rule, measured variables and indices, exclusion and missing data rules decided in advance, the exact confirmatory models and inference criteria, and a short AsPredicted version for co-authors.
0Reviews
Prompt
Act as a research methods consultant who reviews preregistrations for psychology and behavioral science labs and catches the vague lines that later become researcher degrees of freedom. You write in the section order of the OSF Preregistration template and every rule you write could be checked by a stranger with the data. Inputs: - Research question and theory in a few sentences: [ResearchQuestion] - Design: conditions, between or within, randomization, blinding: [DesignPlan] - Participants, recruitment platform, and eligibility: [Sample] - Power analysis inputs (effect size and its source, alpha, power, test): [PowerInputs] - Measures with items, scales, and scoring: [Measures] - Planned analyses as the PI described them: [AnalysisNotes] - Data status (no data, pilot only, existing data accessed or not): [DataStatus] - Output format: [Format] Generate: 1. Study information: title, and hypotheses written as numbered directional predictions (H1, H2) that name the variables and the expected sign. 2. Design plan: study type, blinding, study design, and randomization as OSF asks for them, using DesignPlan. 3. Sampling plan: existing data statement from DataStatus, data collection procedures, sample size, sample size rationale with the power calculation spelled out from PowerInputs, and a stopping rule. 4. Variables: manipulated variables, measured variables, and indices with the exact formula (for example mean of items 1 to 6 with items 2 and 5 reverse scored). 5. Analysis plan: one confirmatory model per hypothesis with the outcome, predictors, covariates, the test, the inference criterion, and what result would count as support; transformations; data exclusion rules (attention checks, completion time cutoffs, outliers) fixed in advance; missing data handling; and a list of exploratory analyses labelled as exploratory. 6. Vagueness audit: quote every line in AnalysisNotes that leaves a choice open (we may control for, outliers will be handled) and the fixed rule that replaces it. 7. A short AsPredicted style summary of the same plan for co-authors who want one page. Constraints: - Do not invent an effect size or cite a paper that was not given; mark NEED SOURCE. - Keep confirmatory and exploratory clearly separate. - No em dashes.
Instructions
Replace every [bracket] with your details before running. Works on ChatGPT, Claude, and Gemini.
Generated Output
This image was generated using the prompt above.
Examples
Example Input
ResearchQuestion: does a 3 minute self affirmation writing task reduce defensive reactions to a health message about sugary drinks among regular consumers DesignPlan: two conditions, affirmation vs control writing, between subjects, randomized by Qualtrics randomizer, participants blind to hypothesis Sample: US adults on Prolific who drink 5 or more sugary drinks a week, 18 to 65 PowerInputs: d = 0.30 from our pilot (n = 80), alpha .05 two tailed, power .90, independent t test Measures: message acceptance, 6 items 1 to 7, items 2 and 5 reverse; intention to cut down, 3 items 1 to 7; attention check 'select strongly disagree' AnalysisNotes: t test on acceptance, maybe control for baseline consumption, outliers will be handled, we may look at gender DataStatus: pilot collected only, main data not collected Format: OSF sections
Example Output
1. Study information Title: Self affirmation and acceptance of a sugary drink health message among regular consumers H1: Participants in the affirmation condition will report higher message acceptance than participants in the control condition. H2: Participants in the affirmation condition will report higher intention to cut down on sugary drinks than control participants. 2. Design plan Study type: experiment. Blinding: participants are not told the hypotheses or which condition they are in. Design: two group between subjects (affirmation vs control writing). Randomization: Qualtrics randomizer with evenly present elements, 1:1. 3. Sampling plan Existing data: registration prior to creation of data; only the pilot (n = 80) has been collected and it will not be included. Procedure: Prolific, US adults 18 to 65 who report 5 or more sugary drinks per week on the screener. Sample size: 470 analyzable participants (235 per group). Rationale: two tailed independent t test, d = 0.30 from our pilot, alpha .05, power .90 gives 235 per group. The pilot estimate is uncertain; this is noted as a limitation. NEED SOURCE if a published estimate should be used instead. Stopping rule: recruit 520 to allow for exclusions; stop when 470 participants pass exclusions or recruitment reaches 600, whichever comes first. No interim analyses. 4. Variables Manipulated: writing condition (affirmation = 1, control = 0). Measured: message acceptance (6 items, 1 to 7); intention (3 items, 1 to 7); weekly sugary drinks (screener). Indices: Acceptance = mean of items 1 to 6 with items 2 and 5 reverse scored (8 minus score). Intention = mean of 3 items. 5. Analysis plan H1: Welch independent t test, acceptance by condition. Support: p < .05 two tailed with the affirmation mean higher. H2: same test on intention. Covariates: none in confirmatory models. A robustness model adds weekly consumption as a covariate and is reported as secondary. Exclusions: fail the attention check; complete in under 40 percent of the median completion time; writing response under 15 words. No outlier removal on outcome scores. Missing data: participants missing any acceptance item are excluded from H1 only. Exploratory: condition by gender interaction; correlation of acceptance and intention. Labelled exploratory in the paper. 6. Vagueness audit - 'maybe control for baseline consumption' becomes: no covariate in confirmatory tests; one prespecified robustness model. - 'outliers will be handled' becomes: no outlier exclusion on outcomes; only the three rules above. - 'we may look at gender' becomes: exploratory interaction, not used to support H1 or H2. 7. AsPredicted summary Data collected? No. Hypothesis: affirmation raises acceptance (H1) and intention (H2). DV: acceptance index; intention index. Conditions: affirmation vs control writing. Analysis: Welch t tests. Exclusions: attention check, speed, short writing. Sample: 470 after exclusions. Other: gender interaction exploratory.