How to Write a Preregistration for a Survey Study (2026)

To write a preregistration for a survey study, you fix the research question, sampling plan, measures, hypotheses, and statistical analysis plan in a public, time-stamped registry before any respondent sees your questionnaire. Most students can draft one in an afternoon once the design decisions are settled. The hard part is not filling in the form; it is deciding what your analysis will look like before you have seen a single response.

This guide walks through the whole process for a questionnaire-based study: which template to pick, what each section should say, the survey-specific details most generic advice skips, and how to report it. It is written for thesis chapters, journal articles, and grant-funded projects in psychology, education, health, and the social sciences.

Table of Contents

1. 1What You Need
2. 2Step-by-Step
3. 3Step 1: Choose the Right Preregistration Template
4. 4Step 2: State the Research Question and Study Design
5. 5Step 3: Define the Population, Sampling Plan, and Target Sample Size
6. 6Step 4: Describe the Survey Materials and Measures
7. 7Step 5: Predefine Hypotheses and Variables
8. 8Step 6: Write the Statistical Analysis Plan
9. 9Step 7: Address Ethics, Data Management, and Transparency
10. 10Step 8: Final Review – How to Write a Preregistration for a Survey Study
11. 11Survey Details Generic Guides Skip
12. 12Preregistration vs Registered Report vs Pilot
13. 13Common Mistakes
14. 14Frequently Asked Questions
15. 15When should I preregister a survey study?
16. 16Do I need a preregistration for an exploratory survey?
17. 17What should I include in a survey preregistration?
18. 18Can I change my preregistration after collecting data?
19. 19How detailed should the statistical analysis plan be?
20. 20Is a preregistration the same as ethics approval?
21. 21Conclusion

What You Need

You cannot write a useful preregistration from a blank form. Every section is a translation of a decision you have already made, or should make first. Before you open a registry, sort out these nine items.

1. A single primary research question. One sentence, phrased so a specific dataset could answer it.
2. A preliminary literature review. Enough to justify your effect size estimate and to show the question has not already been settled.
3. A defined target population. Who you want to describe, bounded by geography, age, and any other criteria that matter to the question.
4. A draft survey instrument. Section headings, item wording, and response scales, even if items are still being revised.
5. A sampling and recruitment plan. The frame, the channel, the target sample size, and how you will handle quotas or screening.
6. Planned measures. Which items form each index, how reverse scoring works, and what reliability or factor structure you expect.
7. Analysis ideas in writing. The specific comparisons you intend to run, not a vague intention to “use regression.”
8. Ethics requirements. Your institution’s review board, informed consent language, incentive level, and data storage plan.
9. The format you need to submit in. Your target journal or funder may name a specific registry and template.

If you cannot write items 3 through 7 without hedging, you are not ready to register yet. Work through how to define operational definitions and how to report survey reliability coefficients before you lock anything in.

Step-by-Step

Eight steps take you from a survey idea to a frozen registration. Each one has a check that tells you it worked, so you can stop when something looks wrong instead of at the end.

Step 1: Choose the Right Preregistration Template

Start at the registry, not the document. If your journal says “we require a preregistration,” its author guidelines usually name a registry too. Many psychology journals ask for an OSF registration using the “General Psychology” or a stage-2 template; economics and political science often point to AsPredicted; systematic reviews belong on PROSPERO; anything involving an intervention on human subjects may need a clinical trial registry.

A mismatch between template and study type is the most common structural error I see. An experimental template with manipulation and control conditions will push you to invent a manipulation your survey does not have, and reviewers notice the padding. Pick the design template that matches a cross-sectional questionnaire, not the one that sounds most rigorous.

Check: the template’s section headings all map onto something real in your study. Any heading you would have to leave blank is a sign you chose wrong.

Step 2: State the Research Question and Study Design

Write one primary question and at most three secondary questions. Flag each as confirmatory or exploratory at this stage, because that label follows it into the results section. A question like “Is there a relationship between perceived stress and job satisfaction among nurses?” is specific enough to analyze. “How do people feel about work?” is a topic, not a question.

Then name the design in plain words: cross-sectional survey, longitudinal panel with two or more waves, experimental survey where an independent variable is manipulated within the questionnaire, or an observational study using secondary data. Survey studies are usually cross-sectional, and saying so removes ambiguity about what your single time point can and cannot support.

Check: someone outside your field could repeat the question back to you accurately. If they would add or drop a word, tighten it.

Step 3: Define the Population, Sampling Plan, and Target Sample Size

This is the section where survey preregistrations beat generic ones. Name the target population first, then be honest about how you will reach it. A probability sample, a student convenience sample, and a paid online panel are three different inference claims, and a reader cannot evaluate your results without knowing which one you made.

State the sampling frame or recruitment channel concretely: a departmental mailing list, a national voter file, a panel provider, a hospital clinic’s appointment list. List the inclusion criteria and the exclusion criteria separately, and include the mechanical ones online samples produce, such as failing an attention check, duplicate responses, or implausibly fast completion times.

Justify the target sample size with a power analysis. Give the effect size you are powered to detect, the alpha level, the test you plan to use, and the resulting minimum N. If you will drop participants under exclusion rules, inflate the recruited target so the analyzed sample still reaches that number. Then set a stopping rule: usually a fixed target N, occasionally a quota structure or a pre-planned sequential design.

Check: your analysis plan cannot contain a subgroup comparison your target N has no power to support. Fix the sample size or cut the analysis.

Step 4: Describe the Survey Materials and Measures

Include the instrument or an appendix link to it, plus the scoring rules. State which items belong to each scale, how many points the response scale has, its anchors, and whether any items are reverse-scored and how. Pre-specify index construction: whether you will delete items that reduce reliability, what your Cronbach’s alpha threshold is, and whether you intend confirmatory factor analysis with a stated factor structure.

Also record the survey’s order, any attention checks, and the completion rules. If you ran a pilot, say what you changed afterwards and that the final version is the one being registered. Report the estimated completion time, since respondents self-select partly on that and a five-minute and a twenty-five-minute survey produce different response patterns.

Check: could a different researcher rebuild each variable from the data without guessing your scoring code?

Step 5: Predefine Hypotheses and Variables

Name each independent variable, dependent variable, mediator, and covariate, then give its operational definition in terms of the actual items. Write hypotheses that can fail. “Participants with higher perceived stress scores will report lower job satisfaction scores, after controlling for years of experience” can come out false. “Stress relates to satisfaction in some way” cannot.

Decide now which analyses are confirmatory and which are exploratory, and write that into each entry rather than sorting it out when you see the p-values. Most templates have an explicit exploratory section; use it instead of smuggling unplanned tests into a confirmatory list.

Plan for missing data. State the threshold for treating a participant or a scale as unusable, whether you will use complete-case analysis, mean imputation, or multiple imputation, and which variables count as missing at item level. Item nonresponse is normal in survey data and you should not decide what to do about it later.

Check: every hypothesis has a named statistical test attached to it somewhere in the document.

Step 6: Write the Statistical Analysis Plan

For each hypothesis, name the test or model: independent-samples t test, chi-square, one-way ANOVA, ordinal regression, multiple linear regression, mixed-effects models for panel data. Include covariates and the rationale for including them, any interaction terms you will test, and whether tests are one-tailed or two-tailed. State your significance threshold and how you will handle multiple comparisons, whether that is a correction or a clearly separated family of confirmatory tests.

Describe the assumption checks you will run and what you will do if an assumption fails, for instance what happens if a planned t test turns out to have badly unequal variances. Set a stopping rule if your design is sequential. End with reporting commitments: which effect sizes and confidence intervals accompany which tests, and which tables appear in the paper.

Keep a plan specific enough that someone could run it. “We will control for age and tenure using multiple regression” is useful. “We will run the appropriate regression” is not.

Check: hand the plan to someone and ask whether they could reproduce your intended output table without asking you a question.

Step 7: Address Ethics, Data Management, and Transparency

Describe the informed consent process, the review board that approved or will approve the protocol, how confidentiality and anonymity are protected, what data will be retained and for how long, and how the incentive works. If the study is a pilot or an amendment of an existing protocol, say which version is covered.

Disclose conflicts of interest and any prior related work, including a pilot on the same instrument. Note what you can and cannot commit to regarding raw data: participant-level survey responses often cannot be shared freely, and saying so up front is better than promising an open dataset you cannot deliver.

Be clear-eyed about limits. A preregistration locks in a plan; it does not guarantee the plan was correct, make a non-representative sample representative, or substitute for ethics approval.

Check: your consent text and your incentive amount match what you wrote in the registration, with no leftover drafts.

Step 8: Final Review – How to Write a Preregistration for a Survey Study

Read the document once as a skeptical reviewer. Every number that appears twice should agree: sample size in the sampling section against the power analysis, exclusion rules against the missing-data plan, hypotheses against the listed analyses. Mismatches there are the fastest way to get a desk rejection on a registered submission.

Check authorship and PI responsibility, attach the survey instrument or supplementary materials, and confirm your target registry’s requirements, since some require a DOI and others do not. Freeze the registration when you are satisfied. Freezing assigns the immutable time stamp and usually a DOI, and the record cannot be edited afterwards; later changes are documented as deviations rather than edits.

Decide on access. Public means anyone can see your hypotheses before you publish, which some teams prefer for credibility. Embargoed means the record is sealed until you publish, and you share an anonymous private link with reviewers. Choose deliberately and note the choice in your methods section.

Check: the registration is frozen and dated before the survey link goes live. That order is the whole point.

Survey Details Generic Guides Skip

Most preregistration advice on the web was written for laboratory experiments. Four pieces of survey-specific detail decide whether your registration is credible.

Expected response rate. State what you expect from your recruitment channel and what you will do if you fall short: extend the field window, add a second channel, or report the achieved response rate as a limitation.

Nonresponse and coverage. Note that a convenience or panel sample limits generalization, and name which claims your sample can support. Comparing a panel to a probability benchmark and reporting weighted and unweighted results is a decision you can make now and regret later.

Panel mechanics. Attention checks, bot filtering, duplicate detection, and quota screens belong in the registration, with the exclusion thresholds spelled out. These are decisions that look arbitrary unless you predeclare them.

Pilot handling. If you piloted the instrument, register the final protocol and note the pilot’s purpose and changes. Pilot data is usually exploratory and normally gets excluded from confirmatory tests.

Preregistration vs Registered Report vs Pilot

Three routes get confused with each other, and they cost very different amounts of time.

Preregistration. No peer review and no in-principle acceptance. It works for any survey study, and it is the right pick when your timeline is short.

Registered report. Peer review happens in two stages, and the journal commits to in-principle acceptance before data collection. It suits high-stakes confirmatory work in journals that offer the format.

Pilot study. No peer review and no in-principle acceptance. Its job is testing an instrument or a workflow, not producing confirmatory evidence.

A registered report moves your design through peer review before data collection, and the journal commits to publication if you follow the accepted plan. That takes months. A plain preregistration is faster and still earns most of the transparency benefit. See our guide to reporting results for how to document each route properly.

Common Mistakes

Registering after the data are in. Some teams collect first and register afterwards, hoping the documentation reads as pre-planning. The timestamp does not lie, and reviewers can tell. It should be preregistered before the link is shared.

Hypotheses that cannot fail. “Participants will report varied attitudes toward the policy” fits any dataset. Fix it by naming variables and a predicted direction or an explicit null hypothesis.

Changing hypotheses once results exist. New questions are allowed, and reporting them as exploratory is standard practice across fields. The failure is an undeclared swap of a confirmatory hypothesis for a favourable one.

Overpromising analyses. Twenty planned comparisons on a sample of 120 will not survive review. Cut to the tests your power analysis supports, and treat the rest as exploratory.

Copying another study’s design section. Lifted text about manipulation and control conditions describes a study you are not running. Write your own, even if it is shorter.

Failing to separate confirmatory from exploratory work. Listing every analysis you might run as confirmatory makes the label meaningless and gives a reviewer grounds to treat all of it as exploratory.

Frequently Asked Questions

Conclusion

Open a plain text document and write four things on one page: the research question, the sampling plan with its target sample size, the measures and how they are scored, and the analysis you will run. Once those exist as sentences, filling in a registry template takes an afternoon. Freeze it before the survey link goes live, then report what you did and what you changed as clearly as you reported the results.