PhD SOP: Explain Your Secondary Data Analysis

PhD SOP: Explain Your Secondary Data Analysis

Using data collected by someone else does not erase your research contribution. Explain the question you investigated, the analysis decisions you owned and the limits of the data. Credit the original collection and distinguish an independent analysis from following an assigned procedure.

Our recommendation is to make the new reasoning visible before defending the fact that the dataset already existed. For a course exercise rather than an investigation, start with the SOP without formal research experience guide. “Secondary data” describes where the material came from. It does not, by itself, establish how much intellectual work your project required or whether your application meets a program's research expectations.

The useful question is what you did with the data and why that work mattered. A sophisticated analysis can be weakly described; a routine exercise can be overstated. The statement should help a reader tell the difference.

Separate four contributions that are often collapsed

A project can involve data collection, curation, analysis and interpretation. The same person may do all four, or different people may be responsible for each. Describe the actual division.

NISO's CRediT descriptions distinguish data curation and formal analysis from other contributions. That vocabulary is useful for clarifying an account, but the taxonomy does not determine authorship or admissions eligibility. You still need to identify your actual work and follow the target's instructions.

Use this inventory before drafting:

The inventory can reveal that your contribution was substantial even without collecting a single observation. It can also reveal that you implemented a supervisor's analysis rather than designed it. Either account can be useful; they should not be written as the same achievement.

Do not claim the dataset answers a question it cannot answer

Existing data arrive with a history. The observations may have been collected for another purpose, from a particular population, with variables that only approximate what you want to study. In the SOP, one relevant limitation can show more judgment than a list of model names.

For example, if you analyzed a convenience sample, do not describe the result as a conclusion about everyone in the country. If the dataset contains an association, do not automatically write that you identified a cause. Discuss the interpretation with the supervisor or collaborator who knows the design.

A statement paragraph is not a methods chapter. Select the limitation that shaped an actual decision: narrowing the question, changing a comparison, checking sensitivity or declining to make a claim. “The data had limitations” is too general to show what you understood.

The negative-results guide addresses cases where the analysis did not support the expected conclusion. The research still in progress guide covers an analysis that has not yet produced a defensible result.

Case A: a new question asked of an existing dataset

GradPilot constructed case; not a real study or applicant. An applicant used an existing education dataset to investigate whether an apparent difference in participation persisted when the analysis distinguished several types of institutions. The data were collected by another organization. The applicant proposed the comparison under supervision and tested alternative groupings.

A possible account is:

Using an existing education dataset, I investigated whether the aggregate participation pattern remained similar across institution types. I proposed separating the categories that the initial analysis combined and compared the resulting estimates with my supervisor. The pattern changed under that distinction, which made me cautious about treating the aggregate result as a description of every institution. I want to study how measurement and grouping decisions shape conclusions drawn from administrative data.

This passage identifies the data source category without claiming collection. It shows a decision, a comparison and a bounded interpretation. Its value does not depend on pretending that the applicant gathered a new national dataset.

The complete statement should name the actual dataset where appropriate and permitted, explain the applicant's precise role and avoid presenting the constructed result as a template to imitate. If the supervisor proposed the grouping, the applicant must credit that fact and focus on the work they did own.

A reader should be able to ask why the categories were selected, what alternative groupings were considered and what the dataset could not establish. Those are substantive questions the paragraph invites.

Case B: an assigned analysis that builds preparation

Second constructed case. A student completed a course project using a public dataset and an instructor-specified method. They learned to prepare the data and interpret the output, but did not formulate the research question or design an independent investigation.

Our recommendation is to describe it as coursework with a specific analytical contribution. Do not call it independent research merely because the data were real.

A truthful passage might read:

In a statistics course project, I implemented the assigned model on a public dataset and investigated why the output changed after a missing-data filter was applied. The question and primary method were set by the instructor. My contribution was to trace the change to which observations remained in the analysis and explain that difference in the report. This experience made me interested in the relationship between data preparation and statistical interpretation.

This can demonstrate preparation and a developing question. It does not establish that the student has independently designed a full research study. For an application that asks for research months or supervised appointments, follow its definitions rather than transferring the project into a category that sounds stronger.

The SOP without formal research experience guide addresses how to use coursework and other preparation honestly. A strong explanation of a bounded exercise is better than a grand description that cannot survive a factual question.

Keep access and research permission separate from writing quality

Do not assume that data being available to your team means they can be shared externally. Human-subject, contractual or repository restrictions may govern access and use. Follow your institution's process before beginning the analysis or sharing material.

The University of Northern Iowa's guidance on existing data distinguishes publicly available, de-identified datasets from privately held data and describes review and consent considerations. Its procedures are institution-specific. The relevant lesson for an application is to report your project's actual authorization accurately, not to infer an exemption from a short web description.

You generally do not need to fill an SOP paragraph with compliance terminology unless it matters to the contribution or the prompt. You do need to avoid claiming approval that was not obtained, implying that you personally collected protected records, or attaching restricted data to demonstrate technical competence.

If permission is unresolved, ask the responsible supervisor or office. A statement review cannot settle that question.

Build the paragraph around a decision

Try this drafting procedure:

1. Write the research question without naming a model.
2. State the dataset's origin and the part of the work you owned.
3. Select one consequential decision, such as an inclusion rule or comparison.
4. Explain how you checked the decision or interpreted its effect.
5. State the result only as broadly as the evidence allows.
6. Connect the remaining question to the training you seek.

Then remove details that merely repeat the CV. A skills list can say that you used R or Python. The SOP should explain why your use of an analysis mattered for the question.

If no decision was yours, the paragraph can still describe learning and execution. Make the boundary clear and consider whether another experience better supports the central research claim.

Do I need to apologize for not collecting the data?

No. Credit the collection and explain your contribution. An apology wastes space without clarifying the work. The important distinction is between the contributions you made and those made by others.

Can I say my analysis was original?

Only when you can explain what was original and have checked that account with the relevant supervisor or literature. A new combination of existing tools is not automatically a new method. Describe the specific question or comparison rather than using “original” as a substitute for explanation.

Review the contribution and its limits

Ask a reader to identify who collected the data, who chose the analysis and what you personally concluded. If those answers are ambiguous, revise the attribution before improving the style.

The research-focused PhD SOP rubric, available through PhD statement review, can help assess the written account. It does not validate the dataset, analysis or research permissions. Find related guidance in the PhD essays hub.

Primary sources checked September 14, 2026. Cases and the drafting procedure are original GradPilot illustrations.