ACADEMIC WRITING GUIDE

How to Preregister a Dissertation Study (With a Complete Worked Example)

By Acadelyra Editorial TeamPublished

You have developed a dissertation research question, chosen a methodology and prepared an analysis plan. You may even have decided which statistical tests you intend to use.

But how will a reader know which decisions you made before seeing the findings—and which decisions were influenced by the results?

Preregistration helps answer that question.

It creates a dated record of your intended research design and analytical decisions before the relevant results are known. Rather than presenting every decision as though it was planned from the beginning, you can show which predictions were specified in advance, which procedures changed and which additional analyses emerged during the study.

This guide explains how to preregister a dissertation study, what to include in the plan, how to use a research registry and how to report deviations honestly. A complete hypothetical university-library study provides a practical example you can adapt to your own methodology.

Important: The study, participants, instruments, numerical illustrations and sample reporting passages in this article are fictional teaching examples. Preregistration does not replace ethics approval, supervisor approval or institutional requirements.

What is preregistration in research?

Preregistration means documenting important aspects of a study before the researcher examines the results that could influence those decisions, then preserving the plan in a dated record.

A meaningful preregistration may describe:

  • Research questions and hypotheses.

  • Study design and recruitment.

  • Inclusion and exclusion criteria.

  • Sample-size decisions.

  • Variables and outcomes.

  • Data collection and management.

  • Planned analyses.

  • Missing-data procedures.

  • Conditions for changing an analytical approach.

  • How deviations will be disclosed.

The objective is not to predict everything that will happen. It is to make the researcher's analytical intentions transparent.

Nosek and colleagues' article, The preregistration revolution, explains the value of distinguishing analyses designed to test prior predictions from analyses developed after observing the data.

A detailed plan can also expose unresolved design problems before they become difficult to fix.

Is preregistration required for a dissertation?

Not universally.

Whether it is required depends on your university, department, discipline, research design, supervisor, funder and applicable research regulations.

A department may require a protocol or approved proposal without requiring a public registry submission. Some journals and research communities encourage or require particular registration practices. Certain regulated studies, including applicable clinical trials, have separate registration obligations that an ordinary dissertation preregistration does not replace.

Check the rules governing your study before selecting a platform.

Preregistration is most useful when it meaningfully clarifies which methodological decisions preceded the relevant evidence.

For a conventional quantitative dissertation, a detailed prespecified analysis plan is often especially useful. For qualitative research, suitable prospective documentation may describe sampling, methodological commitments, reflexivity and broad analytical procedures without pretending that themes or interpretations can all be known in advance.

Preregistration, a research proposal and a Registered Report

These documents serve related but different purposes.

Research planning documents and their distinct purposes
DocumentMain purposeKey distinction
Research proposalExplain and justify a proposed study for review or approvalOften submitted to a supervisor, committee or institution
Analysis planSpecify how research data will be processed and analysedMay remain an internal document unless preserved or registered
PreregistrationCreate a dated, preserved record of research intentionsAllows later comparison between the plan and what was actually done
Registered ReportObtain a particular form of journal peer review of a protocol, typically before results are knownMay involve in-principle acceptance under the journal's rules
Ethics applicationObtain appropriate ethical review and permissionsDoes not automatically constitute public preregistration

A strong proposal can provide much of the material for a preregistration, but copying it into a registry does not automatically make the plan sufficiently precise. Our dissertation data analysis plan guide explains how to organise the analytical decisions before preserving them in a registry.

For proposal structure, see our guide to writing a dissertation research proposal.

When should you preregister your study?

For a prospective quantitative study, the clearest approach is to preregister before collecting or examining outcome data.

The crucial question is whether the decisions documented in the plan could already have been influenced by knowledge of the results.

A plan created after the researcher has reviewed outcome patterns should not be described as though it was registered before those patterns were known.

Existing-data studies require particular care. You may be able to preregister an analysis of an existing dataset if you transparently disclose what information you already know, which variables or outcomes you have inspected, and which decisions remain genuinely prospective.

The Open Science Framework registration guidance includes templates for general quantitative studies, primarily qualitative work and secondary-data analyses.

Before submitting, confirm that the planned timeline and disclosure arrangements suit your project.

Step 1: State the research question precisely

A preregistration begins with the question the study is designed to answer.

A weak example is:

Do reminders help university students?

This does not identify the intervention, target group or outcome.

A stronger question is:

Among eligible students who register for a university library's data-skills workshop, does receiving one additional reminder email increase verified attendance compared with receiving only the usual registration confirmation?

The revised question identifies:

  • The target group: eligible registered students.

  • The intervention: one additional reminder email.

  • The comparison: the usual confirmation without that extra reminder.

  • The outcome: verified workshop attendance.

  • The relevant setting: a university library workshop.

The question should be aligned with your approved research design rather than chosen simply because the outcome is easy to measure.

Our research-hypothesis guide explains how to convert a research question into a measurable prediction.

Step 2: Specify hypotheses and distinguish primary from secondary questions

For the hypothetical workshop study, the researcher proposes:

Primary directional hypothesis (H1): Students assigned to receive the additional reminder will have a higher probability of verified workshop attendance than students assigned to the usual-confirmation condition.

Null hypothesis (H0): The attendance probability is the same in the two assignment groups.

A prespecified exploratory question asks whether the assignment-group attendance difference varies by registration lead time. The plan will compare students who registered at least seven days before the workshop with those who registered less than seven days before it, using the recorded registration and scheduled-session timestamps. Group counts, denominators and risk differences will be presented descriptively within each category; no subgroup significance tests or confirmatory interaction claims are planned.

That secondary question should not be quietly converted into the primary hypothesis merely because it produces an interesting result.

It is also important to define the direction of the hypothesis separately from the statistical reporting convention. A directional substantive prediction does not automatically require a one-sided significance test.

In this worked example, the planned primary significance test will be two-sided.

Step 3: Define the design and allocation procedure

The fictional study is an individually randomised, parallel-group experiment for one large university-library data-skills workshop. Using one scheduled session keeps the teaching example focused on independent student assignments rather than a design involving multiple workshop clusters.

The researcher proposes to enrol 240 eligible registrations, assigning 120 to each condition if recruitment reaches the target and the allocation remains balanced.

Every student receives the ordinary registration confirmation.

For this teaching protocol, the workshop is scheduled for 1 February 2027 at 10:00 Africa/Nairobi. Recruitment opens on 1 January 2027 at 10:00 and closes at the earlier of the 240th eligible registration or 28 January 2027 at 10:00. These are fictional planning dates. Later registrations and any waiting list are outside the randomised sample. There will be no interim attendance comparison or outcome-dependent extension.

Students assigned to the intervention group additionally receive one reminder 48 hours before the workshop, on 30 January 2027 at 10:00. The approved message repeats the session time, venue and joining instructions; it adds no incentive or additional teaching material.

Students in the comparison group receive no additional research reminder. Other routine communications must be the same across groups, except for the defined intervention.

An independent allocation custodian will create a list containing 120 reminder labels and 120 comparison labels and shuffle it once using a reproducible computer-generated random sequence. Eligible students receive the next unused label in registration order after eligibility is confirmed. If recruitment stops below 240, the realised group totals need not be equal.

The custodian will record the software, version, random-number generator and seed in a restricted allocation log before recruitment. Recruitment staff cannot inspect future assignments; the seed and full sequence will not be disclosed during recruitment. The intervention sender receives only the details needed to deliver the assigned messages. Attendance staff will use a register without assignment labels where practicable. The analysis will record any departures from these arrangements.

This design could support a causal estimate of the assigned reminder's effect in the study setting if randomisation and outcome ascertainment are appropriately implemented and the relevant causal assumptions hold. For example, students forwarding reminders to one another could create spillover between groups. Delivery failures and spillover should be documented; the assignment comparison does not isolate the effect of actually reading the message.

It does not automatically establish effects for all universities, all students or every type of reminder.

Ethics review, consent arrangements and lawful handling of student contact information must be settled before recruitment.

Step 4: Specify eligibility and exclusions before seeing outcomes

Eligibility criteria should follow from the population and design.

For this hypothetical study:

Include: Currently enrolled students aged 18 or over who register within the stated recruitment window, have a working institutional email address and provide the permissions required by the approved study arrangements.

Exclude before randomisation: Registrations outside that window, records not meeting those criteria and registrations cancelled before allocation. Use the approved institutional identifier to resolve duplicates before assignment, retaining each student's earliest valid registration. Each student may enter the randomisation only once.

After randomisation, students should not simply be removed because they did not open the reminder, were absent, reported disappointment or had an inconvenient outcome.

The primary analysis will follow the assignment principle: retain randomised participants in their allocated groups, subject to appropriately specified handling of genuinely missing or unavailable outcome records.

If a participant withdraws consent or data use is restricted, follow the approved consent and ethics requirements. Preregistration cannot override those obligations.

Document the number eligible, randomised, included in analysis and excluded, with reasons at each stage.

Step 5: Justify the sample-size decision

A preregistration should explain how the target sample was chosen.

For the hypothetical example, the target is 240 eligible students.

Suppose the library expects this number to be feasible based on its workshop capacity and prior registration volume.

That would be a feasibility-based target, not evidence that the study is sufficiently powered to detect the predicted effect.

Do not claim that 240 participants provide 80% or 90% power unless an appropriate calculation establishes that result under stated assumptions.

A valid power analysis for a two-group attendance comparison would need inputs such as:

  • Expected attendance probability in the comparison group.

  • The smallest effect worth detecting.

  • Type I error rate.

  • Desired power.

  • Allocation ratio.

  • Anticipated missingness, clustering or other relevant design features.

These assumptions may be uncertain and should be justified from appropriate evidence.

For the calculation and reporting process, see our guide to conducting power analysis for a dissertation.

A feasibility-based sample can still be useful, but its precision and statistical limitations must be acknowledged.

Step 6: Define variables and outcomes unambiguously

The fictional study's primary variables are:

Prespecified variables and outcome measures for the fictional workshop study
ItemPrespecified definitionMeasurement
AssignmentAdditional reminder versus usual confirmationRandomisation record
Primary outcomeWhether the student attended the scheduled workshopVerified attendance record
Analysis unitEligible randomised registration, with each student represented onceApproved participant identifier
Outcome windowThe scheduled workshop session for which the student registeredWorkshop register
Planned effect measureAttendance-probability difference between groupsIntervention proportion minus comparison proportion

Attendance is coded as 1 for confirmed attendance and 0 for confirmed nonattendance.

A missing register entry is not automatically coded as nonattendance if the register is incomplete or unreliable.

For this example, attendance means a match to the student's approved identifier on the session's verified entry register. The attendance team will reconcile scanner records against the staff backup register and close ascertainment seven days after the scheduled session. Nonattendance is coded 0 only if the reconciled register is certified complete and contains no entry for that student. Unresolved identity or register problems are missing outcomes. Attendance at a different session does not count toward this outcome.

A cancellation or transfer requested by a student after assignment does not change the assigned group or original outcome window; if the original session runs and its register is complete, absence from it is nonattendance. A bounced or unopened reminder likewise does not remove a participant. If the entire workshop is cancelled, all attendance outcomes are unavailable, the cancellation is reported and no primary comparison is calculated. There is no replacement session in this teaching protocol.

The primary outcome should be defined before the results are known, not replaced afterward with an alternative such as email opening or workshop satisfaction because it creates a more favourable result.

Step 7: Write the primary analysis plan

A useful preregistration explains not just which software or test will be used but how the data will produce the answer.

The hypothetical study will compare verified attendance probabilities according to randomised assignment.

The primary estimand is the difference between the probability of attendance under assignment to an additional reminder and the probability under assignment to usual confirmation, within the trial population and specified conditions.

Primary estimate:

Attendance proportion in reminder group − attendance proportion in comparison group.

For example, if attendance rates were 0.60 and 0.50, respectively, the risk difference would be:

0.60 − 0.50 = 0.10

That is a difference of 10 percentage points, not a 10% relative increase.

Those two values are imaginary arithmetic examples, not observed study outcomes.

The researcher will report the observed risk difference with a 95% confidence interval using Newcombe's hybrid-score method without continuity correction, combining Wilson score intervals for two independent proportions, together with the group counts and denominators. Newcombe's methodological paper evaluates this interval construction.

The null hypothesis of equal attendance proportions will be evaluated using a two-sided Fisher's exact test at a significance level of 0.05. The implementation will use R's stats::fisher.test with alternative = "two.sided" and simulate.p.value = FALSE; the software version and analysis script will be retained. Its official documentation specifies the probability-ordering convention for the two-sided p-value. The confidence interval returned by that function concerns an odds ratio; it is not the separately calculated risk-difference interval.

A confidence interval for the risk difference and a Fisher exact p-value need not be numerically equivalent inferential procedures. Their intervals and tests should not be represented as interchangeable or as providing an identical threshold decision.

The primary conclusion will consider the estimated magnitude and uncertainty, not just whether the p-value is below 0.05.

The researcher will use the prespecified analysis without silently switching to another method after seeing the results.

Step 8: Plan missing-data handling and sensitivity checks

Even an apparently simple attendance outcome may be incomplete.

A workshop register could be missing because a session was cancelled, a scanner failed or participants attended a transferred session.

The preregistration should state how the researcher will distinguish verified absence from unavailable outcome information.

For this fictional study:

Primary approach: Retain the original assignment labels and analyse all participants whose attendance outcome can be verified. In each group, divide verified attendances by the number with verified outcomes. This is a prespecified complete-outcome analysis, not an analysis that supplies outcomes for every randomised student. Report the randomised total, observed-outcome denominator and missing count separately for each group.

If either group has no verified outcomes, do not calculate the primary risk difference, confidence interval or Fisher test. State that the planned comparison could not be estimated. Otherwise apply the primary methods already specified, even when some outcomes are missing.

This analysis may differ from the intended effect in all randomised students: if missingness is related to assignment or attendance, analysing only observed outcomes can undermine the protection provided by randomisation. It is not automatically unbiased, and no missingness test can certify away that concern.

Prespecified sensitivity scenarios: For participants whose data may lawfully be analysed, let N be the total in an assignment group, A the verified attendances and M the unavailable outcomes. Recalculate the reminder-minus-comparison risk difference under all four group-level assignments of the missing binary outcomes: all missing outcomes equal 0 in both groups; all equal 1 in both groups; reminder missing outcomes equal 0 and comparison missing outcomes equal 1; and the reverse. In each scenario, the group attendance proportion is either A / N or (A + M) / N. The last two scenarios give lower and upper bounds on the sample risk difference under these binary assignments.

These are deterministic sensitivity scenarios, not recovered observations or confidence intervals. Do not attach a new significance claim to a preferred scenario. They do not solve selection bias or identify the population effect. Records that cannot be used because consent was withdrawn or access is prohibited are not silently reconstructed; report the permitted participant flow and explain that those restrictions limit the scope of the analysis and scenarios.

The primary rule and these scenarios are fixed before inspecting outcomes. A more complex real study may require a different, justified missing-data strategy. For more guidance on checking the dependence of conclusions on assumptions, see our sensitivity-analysis article.

Step 9: Define conditional decisions before analysis

A preregistration does not require researchers to ignore unexpected technical problems.

It can specify decisions that will be made if a defined problem occurs.

Examples include:

Conditional decisions specified before outcome inspection
SituationPrespecified response
Entire workshop cancelled before deliveryRecord all outcomes as unavailable, report the cancellation and do not calculate the primary comparison
Reminder delivery fails for some studentsPreserve original assignment in the primary analysis; describe implementation failure
Attendance record incompleteApply the prespecified missing-outcome procedure and sensitivity scenarios
Fewer students register than expectedReport achieved recruitment and precision; do not extend enrolment selectively according to interim results
Unexpected outcome pattern appearsPreserve the primary analysis and label additional analyses as exploratory

Conditional rules should be specific enough to prevent the researcher from choosing a preferred path after looking at the findings.

The Center for Open Science's preregistration guidance supports specifying conditional decisions and disclosing changes alongside unplanned analyses.

Step 10: Select an appropriate registry and submit the plan

The Open Science Framework (OSF) provides registration templates for different research designs.

For a standard quantitative study, a comprehensive general-purpose preregistration template may be appropriate.

Other projects may need a qualitative, secondary-data or domain-specific form.

A practical process is to:

  1. Check whether your institution, funder, journal or research community requires a particular registry or format.

  2. Select a template suited to the study design.

  3. Prepare and review the full plan before submission.

  4. Check for confidential participant information, identifying details, proprietary content and other material that must not be disclosed.

  5. Submit the registration according to the platform's current instructions.

  6. Retain the registration reference, date and accessible documentation for later reporting.

Research plans can contain information that should not be made public immediately. Investigate appropriate embargo, restricted-access or other institutionally approved arrangements before submitting.

A registered plan is not a substitute for ethics review, approved recruitment procedures or secure data management.

The Center for Open Science's template-selection guide provides additional guidance on choosing the appropriate form.

Complete worked preregistration example

The following worked registration brings the preceding decisions together. Read it with Steps 1–10 for the operational rules. It is a fictional teaching protocol, not a submitted registration or an ethics-approved study.

Study title

Effect of one additional email reminder on attendance at a university library data-skills workshop: A randomised study.

Research question

Among eligible student registrants, does assignment to an additional reminder increase verified attendance compared with usual confirmation alone?

Hypothesis

The additional-reminder assignment group will have a higher workshop-attendance probability than the usual-confirmation group.

Study design

Individually randomised, two-arm study of one workshop scheduled for 1 February 2027 at 10:00 Africa/Nairobi. An independent custodian shuffles 120 labels per group and assigns the next unused label to each eligible registrant, concealing future assignments. The additional reminder is sent 48 hours before the session; usual confirmation and other routine communications are the same in both groups. Allocation software, version and seed are retained in the restricted log described in Step 3.

Population and sampling

Eligible students registering from 1 January 2027 at 10:00 until the earlier of the 240th eligible registration or 28 January 2027 at 10:00, Africa/Nairobi. Target: 240, chosen for feasibility rather than demonstrated statistical power. No replacement participants, outcome-dependent extension or interim attendance analysis. Group totals may differ if recruitment stops early.

Eligibility

Enrolled students aged at least 18 with a working institutional email address and required study permissions. Confirm eligibility and resolve duplicate institutional identifiers before assignment, retaining the earliest valid registration. Exclude registrations outside the recruitment window and those cancelled before allocation.

Primary outcome

Verified entry at the original scheduled session, matched by approved identifier. Reconcile scanner and backup registers within seven days. Code 1 for a verified entry and 0 for absence only from a certified complete register; unresolved records are missing. Other-session attendance does not count. Preserve assignment despite unopened or undelivered reminders and participant cancellations or transfers. If the whole session is cancelled, no primary comparison is calculated.

Primary analysis

Use verified attendances divided by verified outcomes within each original assignment group. Estimate reminder minus comparison attendance probability with a 95% Newcombe hybrid-score interval without continuity correction. Conduct R's two-sided Fisher exact test without simulation at alpha = 0.05. Do not substitute the odds-ratio interval returned by Fisher's test for the risk-difference interval. Retain analysis code and software versions; report group numerators, randomised totals, analysis denominators and uncertainty. If either analysis denominator is zero, report that the primary estimate and test are unavailable.

Missing data

Use the complete-outcome primary analysis just specified. Report missing numbers and reasons by group. Apply all four binary missing-outcome scenarios in Step 8 using A / N or (A + M) / N for permitted records; show both extreme group assignments as bounds, not confidence intervals. Explain selection bias and any consent-related restrictions.

Additional analyses

Prespecified exploratory analysis: present assignment-group counts, denominators and risk differences separately for registration lead times of at least seven days and less than seven days. No subgroup significance tests or confirmatory interaction claims. Any further analysis prompted by the observed results will be identified as post hoc exploratory work and will not replace the primary comparison.

Research ethics and data management

Obtain required institutional approval and permissions before recruitment, including the approved consent and withdrawal arrangements. Keep contact details, the identifier key and the full allocation log in institution-approved restricted storage, separate from the coded analysis data. Give recruitment, reminder-delivery, attendance and analysis staff only their authorised access. The institution's actual storage location, retention period and disposal procedure must be inserted before real registration; this teaching example does not invent institutional authorisation. No participant identifiers, email addresses or allocation seed will be published during recruitment.

Deviations

Maintain a dated record of differences between the registered plan and actual study conduct, with reasons, timing and implications.

Registration status

Not submitted. This example has no registration identifier, ethics approval or actual participant data.

This worked model connects the research question, assignment, outcome, stopping rule and analysis. Before actual registration, replace the fictional dates, confirm institutional arrangements and supply the selected registry's required study-specific details. Check that the registration and attached protocol agree; preserve the complete submitted version.

What if you need to change a preregistered plan?

Research rarely follows every planned step perfectly.

A computer system may fail. A recruitment target may prove unattainable. An instrument may become unavailable. A procedure may need adjustment for safety, ethics or feasibility.

The important distinction is between changing a plan and concealing that the plan changed.

Preregistration does not prohibit justified modifications. It creates a record against which those modifications can be understood.

A transparent deviation record can include:

Fields for a transparent preregistration deviation record
FieldWhat to document
Original planThe exact registered decision
Actual changeWhat was done differently
ReasonWhy the adjustment was necessary
TimingWhether the change occurred before or after outcome inspection
ConsequenceHow the modification may affect interpretation
DisclosureWhere the deviation is reported in the dissertation

If the registry supports updates or linked amendments, use the appropriate mechanism while preserving the original record.

Do not overwrite the historical plan and imply that the revised procedure was the original preregistered method.

Worked example: Reporting a deviation honestly

Suppose the hypothetical library planned to recruit 240 students but the workshop generated only 198 eligible registrations before the fixed recruitment deadline.

Assume the researcher learned this after recruitment had closed, without examining attendance differences between the randomised groups.

A defensible deviation statement could read:

The study was preregistered with a feasibility-based recruitment target of 240 students. Recruitment closed with 198 eligible randomised participants because fewer workshop registrations were available during the approved data-collection period. The study did not extend recruitment based on an interim comparison of attendance outcomes. The original primary outcome and analytical approach were retained. The smaller achieved sample may reduce precision and the ability to detect modest differences; this limitation is considered when interpreting the estimates.

This paragraph illustrates reporting style. It does not describe a real study.

If the researcher had examined interim group differences before making the recruitment decision, that fact would require additional disclosure. It could have important consequences for the interpretation of the planned statistical analysis.

How to report preregistration in a dissertation

A preregistration reference can appear in the methodology chapter, with a corresponding discussion of deviations in the results or discussion as appropriate.

Example methodology paragraph

The study protocol and primary analysis plan were preregistered before examination of the relevant outcome data. The registration specified the research question, allocation procedure, primary outcome, analysis population, risk-difference estimator, significance-testing approach and missing-data procedures. The registration identifier and submission date are reported in the dissertation's methods documentation. Any deviations from the registered procedures are identified separately.

Replace the generic registration description with the actual reference and factual timing.

Never claim preregistration if the plan was only written privately after the outcomes were examined.

Example results paragraph

The results are presented according to the prespecified primary comparison between the reminder-assignment and usual-confirmation groups. Group numerators and denominators, the estimated attendance risk difference and associated uncertainty are reported. Any incomplete attendance records and their prespecified sensitivity analyses are also disclosed. Additional analyses are identified as exploratory rather than treated as originally planned hypothesis tests.

This is an example of reporting structure, not actual numerical results.

Example discussion paragraph

Preregistration clarified the primary analytical decisions made before examining the outcome data. Nevertheless, registration does not eliminate the effects of recruitment limitations, incomplete outcome measurement, intervention-delivery problems or possible weaknesses in the design. The findings are therefore interpreted with reference to both the registered plan and the study's observed implementation.

The value of preregistration is greater when the dissertation visibly connects the preserved plan with the analyses actually reported.

Does preregistration prevent exploratory research?

No.

Exploratory research is valuable for identifying unexpected patterns, developing hypotheses and identifying future studies.

The problem arises when a hypothesis developed after examining the data is presented as though it was predicted before the analysis.

A researcher can report:

The preregistered primary comparison examined workshop attendance by assignment. The registration-lead-time comparison was prespecified as exploratory. A separate, additional comparison by students' year of study was developed after the primary results were examined and is reported as post hoc exploratory work.

This is more informative than silently presenting the exploratory subgroup as an original confirmatory prediction.

A preregistration is not proof that the planned methods were appropriate or that the final analysis is unbiased.

Methodological validity still requires suitable measurement, credible design, correct calculations and transparent interpretation.

Preregistration for qualitative dissertations

Qualitative preregistration requires methodological sensitivity.

Some approaches use relatively structured procedures, while others deliberately emphasise flexible interpretation, reflexivity and evolving research questions.

A qualitative preregistration may document:

  • The research purpose and methodological approach.

  • The proposed sampling strategy and recruitment criteria.

  • Intended types of evidence and data collection.

  • Ethical and confidentiality safeguards.

  • The researcher's positionality and relevant assumptions.

  • Planned broad analytical procedures.

  • Circumstances in which sampling or analysis may be adapted.

  • How methodological changes and reflexive decisions will be documented.

Do not promise a fixed number of themes, exact interpretations or mechanical agreement scores when those commitments conflict with the chosen qualitative approach.

OSF includes a qualitative preregistration template, but it is not obligatory for every qualitative study. The template should support—not distort—the methodology.

Common preregistration mistakes

Registering after examining the results without disclosing that timing

A retrospective record may still document procedures, but it cannot establish that the analysis was prespecified before those results were known.

Writing a vague analysis plan

Saying “data will be analysed in statistical software” does not identify the outcome, estimator, analysis sample or decision rules.

Changing the primary outcome after seeing the data

A different outcome may answer a different question. Disclose the change, preserve the original plan and distinguish additional analysis from the planned test.

Treating preregistration as proof of research quality

A preregistered study can still have a biased sample, poor measurement, inappropriate analysis or inaccurate conclusions.

Claiming a feasibility-based sample is statistically powered

A convenient recruitment target is not a substitute for a justified sample-size calculation.

Using an unsuitable template

A form developed for one kind of study may impose assumptions that do not fit another.

Concealing deviations

Unanticipated problems can occur. The correct response is to document them, explain their implications and report the actual procedures transparently.

Publishing confidential research information

A preregistration may contain sensitive recruitment, location or institutional details. Follow applicable ethics, privacy and research-governance requirements before sharing it.

Reporting only favourable planned tests

Preregistration is weakened when a researcher omits unfavourable prespecified outcomes or selectively interprets the analyses.

Frequently asked questions

Can I preregister a master's dissertation?

Yes, where suitable for the research design and permitted by the institution. It is not restricted to doctoral research.

Is preregistration the same as submitting my proposal?

No. A proposal may contain similar information, but preregistration emphasises preserving a dated record of analytical intentions. A proposal submitted to a supervisor is not necessarily a public or formal registration.

Can I preregister a study using existing data?

Potentially. Explain what you already know about the dataset, whether the outcomes or relationships have been inspected and which analytical decisions remain genuinely prospective.

Can I amend a preregistration?

You may document changes and, where supported, create a linked amendment or additional registration. Preserve the original record and describe when and why the modification occurred.

Does preregistration guarantee publication or supervisor approval?

No. It does not guarantee ethical approval, a pass, a favourable result or acceptance by a journal.

Does preregistration mean I cannot explore unexpected findings?

No. You can explore unexpected patterns; report them as exploratory when that accurately reflects their origin.

Do I need to share participant data publicly?

Preregistration does not itself require disclosing confidential participant-level data. Follow the registry's procedures and the privacy, ethics and data-management rules governing the project.

Should I preregister if my dissertation is qualitative?

Consider whether prospective documentation meaningfully improves transparency within your methodology. Do not force a rigid confirmatory statistical framework onto an incompatible qualitative approach.

Checklist before submitting a preregistration

  1. Is the research question specific enough to determine what counts as an answer?

  2. Are the hypotheses and primary outcomes clearly distinguished from exploratory aims?

  3. Does the research design fit the questions?

  4. Are eligibility and exclusion rules specified?

  5. Is the sample-size decision justified without unsupported power claims?

  6. Are variables and measurement procedures operationally defined?

  7. Is the primary target quantity or estimand identified?

  8. Are the primary analysis and uncertainty methods specified?

  9. Is missing-data handling defined before outcome inspection?

  10. Are conditional analytical decisions sufficiently clear?

  11. Does the plan identify foreseeable deviations and how they will be documented?

  12. Are confidentiality and ethics requirements addressed?

  13. Does the selected registry or template fit the study?

  14. Can the future dissertation distinguish planned and exploratory analyses?

  15. Have the supervisor and relevant institutional requirements been checked?

Final takeaway

A strong preregistration makes your research decisions easier to examine.

It explains what you intended to test, how you planned to collect and analyse the evidence, and which decisions were made before the results could influence them.

It should be detailed enough to guide the analysis, flexible enough to accommodate transparently documented real-world problems and honest enough to distinguish planned tests from later discoveries.

Preregistration will not guarantee that a dissertation is correct. But it can make the study's reasoning more transparent, the methodological record easier to evaluate and the final conclusions more accountable to the evidence.

References and further reading

Editorial note: All illustrative study descriptions, numbers, procedures and sample reporting paragraphs are hypothetical. Verify methodological, ethical and registry requirements before applying the guidance to a real project.

Need help developing a defensible dissertation plan? Explore Acadelyra's academic support services for research-planning support, methodological clarity, academic editing and assistance presenting research decisions transparently while respecting academic integrity.