A data analysis plan is a written description, completed before touching the data, of how you will analyze each objective of your thesis: which variables, which test or model, which assumptions you will check, and how you will handle missing values. It may sound bureaucratic, but it is one of the most powerful tools for getting your thesis through without surprises, and committees and supervisors are increasingly asking for it explicitly.
Why they ask for it (and why it benefits you)
Deciding the analysis in advance has three advantages that matter both to the committee and to you. First, transparency: it separates what you planned from what you discovered, and prevents the bias of choosing the test that gives significance after looking at the data (the classic p-hacking that reviewers detect). Second, it saves time: when you have the data, you execute rather than improvise. Third, it is your best methodological defense: a clear plan turns the committee's question "why did you use this analysis?" into an answer you had already written. If you also register it on a platform like OSF (pre-registration), the shield against reviewers is even stronger.
And before you call it finished, run it through the Methodologist: you paste the project or the protocol and it returns the design flaws worth fixing, whether the sample size can carry the model you want to estimate, a read of the analysis plan, and a draft preregistration. All of that is cheap to fix now and extremely expensive to fix once the data are in.
What a complete analysis plan includes
- Objectives and hypotheses, numbered, each with a directional prediction where applicable.
- Variables: dependent, independent and control, with their type and measurement scale.
- Test or model assigned to each objective, justified by the design.
- Assumptions you will check and what alternative you will use if they fail.
- Missing data handling (listwise, multiple imputation) and outlier strategy.
- Sample size and power: a priori justification (G*Power or simulation).
- Significance level and correction for multiple comparisons if applicable.
- Sensitivity analyses planned, if any.
Worked example
Imagine a thesis evaluating whether an intervention program reduces anxiety and whether the effect depends on sex. The analysis plan, in its most useful form, is a table that maps each objective to its test:
| Objective / hypothesis | Variables | Analysis |
|---|---|---|
| Describe the sample | All | Descriptives (M, SD, %) |
| H1: the program reduces anxiety | Anxiety (continuous), group (intervention/control), time (pre/post) | 2x2 mixed ANOVA or mixed model |
| H2: the effect differs by sex | + sex (categorical) | Group x sex interaction in the model |
| H3: greater adherence predicts greater improvement | Adherence and change (continuous) | Correlation / regression |
With that table written, when the data arrive you know exactly what to run. And if you are unsure which test to assign to an objective, the guide on how to choose the statistical test resolves it by design; the full analysis workflow is in how to analyze thesis data step by step.
Sample size belongs in the plan, not afterwards
An analysis plan without a sample size justification is incomplete. Power analysis (how many participants you need to detect the effect you expect) is decided before data collection, not after. You can estimate it with the sample size calculator, and it is worth understanding why post hoc power analysis does not justify anything.
Where it goes in the thesis
The analysis plan lives in the Methods section, in a subsection titled "data analysis" or "statistical analysis plan", typically at the end of the Methods and before the Results. The guide on how to write the Methods section details what to include and in what order for reproducibility.
Quick checklist
Before signing off on your plan, check: every objective has an analysis; every analysis is justified by the variable types and design; you have anticipated what to do if assumptions fail; the sample size is justified a priori; and you have declared how you will handle missing values. If all five are in place, you have a defensible plan. Defensible in the literal sense: a doctoral committee usually starts with the method, so run the plan through the defense simulator and see which questions it raises.
Need a data analysis plan that will hold up to committee scrutiny?
I am a PhD in Psychology and I design the analysis plan for your thesis (objective-to-test map, assumptions, power and missing data handling), ready for your supervisor and committee, in psychology or any health discipline. Free initial diagnosis.
See the statistical consulting →Frequently asked questions
What should a data analysis plan include in a thesis?
A complete data analysis plan should specify: the research questions and hypotheses, each variable with its level of measurement, the primary and secondary analyses, the statistical tests and their justification, how assumptions will be checked, how missing data will be handled, the significance level (alpha), and the effect size indices to be reported.
Why should I write a data analysis plan before collecting data?
Pre-specifying your analysis prevents p-hacking and HARKing (Hypothesizing After Results are Known), which inflate false positive rates. It forces you to think through your design and analysis strategy before seeing the data, and gives your committee and reviewers confidence that the findings are genuinely confirmatory.
How does a data analysis plan relate to pre-registration?
A data analysis plan is the core document of a pre-registration. Registering your plan on OSF or AsPredicted before data collection creates a time-stamped public record of your hypotheses and analysis strategy, which you can cite in your manuscript to clearly distinguish confirmatory from exploratory findings.