ACADEMIC SUPPORT · UAE & GCC
Dissertation Data Analysis Plan: Map Questions to Variables and Tests
Plan how each question will be answered before collecting a dataset that cannot answer it.

A dissertation data analysis plan is a short, reviewable map from each research question to the data required, the variables or codes, cleaning rules, descriptive summaries, primary analysis, assumptions and presentation. Draft it before data collection where feasible. State how missing responses and departures from the plan will be handled. Do not choose a statistical test simply because it is available in SPSS, or promise findings the design cannot support.
The US National Center for Advancing Translational Sciences defines an analysis plan as a roadmap for organising, analysing and presenting data and recommends establishing it when planning the study. Your programme handbook, ethics approval and supervisor determine the detail required. A student project needs a plan appropriate to its question; it does not need to imitate every section of a clinical trial statistical analysis plan.
This guide gives a practical workflow for UAE and GCC postgraduate researchers using their own lawful data. Labeeb’s Academic data-analysis support is advisory and analytical within an agreed scope. The student remains the researcher, author and owner of the design, data and interpretation.
Why plan before you collect?
A pre-collection plan exposes gaps while they are still fixable. A questionnaire may lack the item needed for the primary outcome; a planned group comparison may have too few cases in one group; an interview prompt may not address the stated question. The Canadian Journal of Hospital Pharmacy research primer explains that analysis planning should guide a quantitative study from summaries through hypothesis testing before data collection. Its examples are health research, so adapt the principle, not a specific medical test, to your dissertation.
Question
State the primary question and define what an answer would look like. Separate planned analyses from optional exploration.
Data
Specify the source, participants, measurement timing and variable definitions, including lawful access and consent.
Decision
Choose summaries and analysis that fit the design, then decide how uncertainty and limitations will be reported.
A plan is not a promise that all assumptions will hold. It should say what will be checked and how the team will document a change. NCBI’s research-data stewardship chapter lists population, datasets, variables, missing values and analysis order among the items to address.
Build a question-to-analysis matrix
Use one row per primary question or hypothesis. This forces the proposed evidence and analysis into the same view. The example below is fictional: a UAE university surveys full-time students about commute duration and class attendance. The study is observational; an association cannot establish that commuting causes absence.
| Question or task | Required data | Planned analysis | Interpretation limit |
|---|---|---|---|
| What is the distribution of commute duration? | Self-reported minutes and a defined recall period | Count, median, spread and a distribution plot | Self-report error; one campus is not the whole UAE. |
| Is commute duration associated with attendance? | Commute minutes and a lawfully accessible attendance percentage for the same period | Plot and an association model chosen after variable/design review; report estimate and interval | Other factors may explain the relationship; no causal claim. |
| Are responses missing systematically? | Item completion, participant group and recruitment channel | Summarise missingness and compare available characteristics where permitted | Missing data may bias the final estimate. |
The matrix does not prescribe a particular correlation or regression test. That decision depends on the actual variable scales, design, distribution, confounders and supervisor-approved question. If attendance records cannot be lawfully accessed, revise the question and measure before recruitment; do not invent a proxy after collecting responses.
Write the data dictionary and cleaning rules
A dataset is not ready merely because it opens in SPSS or Excel. Define each variable’s name, meaning, allowed values, units, missing code and source. Record whether a scale score is the sum or mean of particular items, whether any items reverse-code, and what will happen if one item is blank. Keep the original raw file read-only and create a versioned analysis copy. Never alter a participant’s answer to improve a result.
| Planning field | Illustrative entry | Check before analysis |
|---|---|---|
| commute_minutes | Self-reported one-way minutes; integer 0–240 expected | Flag implausible values for source check rather than silently deleting. |
| attendance_pct | Recorded scheduled-class attendance for the study period | Confirm permission, denominator and matching period. |
| missing value | Blank means no answer; “not applicable” has a separate code | Do not interpret blank as zero. |
Labeeb’s existing SPSS data-cleaning checklist addresses the operational checks after data arrive. This article covers the earlier design decision: what should be collected and how it will answer the question. The two tasks should not be confused.
Specify the primary analysis and its assumptions
For every quantitative question, describe the outcome and predictor or groups, the proposed summary, the planned model or comparison and the diagnostics. Include the size and direction of an effect with uncertainty, not only a p-value. State any planned adjustments and why those variables were selected. Do not test many outcomes and report only the one with the smallest p-value as though it was pre-planned.
The statistical principles for prospective protocols recommend deciding methods before collection and keeping the final report tied to the stated analysis. A recent observational-study analysis-plan template offers more detail for complex projects, but it is a specialist reference, not a universal university template.
Qualitative analysis also needs a plan: which material will be included, how it will be transcribed or prepared, what coding approach is intended, how analytic decisions will be documented, and how participant confidentiality will be preserved. Do not present a qualitative coding decision as if it were a statistical significance test. Mixed-methods work needs both strands and an explicit integration point.
Plan missing data and honest changes
Write down how missing cases will be counted and described, which variables are essential for the primary analysis, and what conditions would prompt a different method. The correct response to unexpected missingness depends on why data are missing, the extent of the problem and the study design. Do not promise that a software default will solve it. A statistician or supervisor may need to approve sensitivity analyses.
“The primary association uses participants with valid commute and attendance data. We will report how many were excluded and compare available characteristics between included and excluded participants where permission and sample size allow. If the planned model assumptions fail, we will document the diagnostic and a supervisor-approved alternative, distinguishing it from the pre-specified analysis.”
This language is an example of transparency, not a method to copy blindly. A legitimate protocol amendment should be dated, explained and reflected in the ethics record if it changes consent, data access or participant burden. A result-driven change hidden from readers undermines trust.
Turn the plan into a results structure
List the tables and figures you expect before seeing outcomes: participant flow and exclusions, descriptive characteristics, the primary estimate with uncertainty, and any planned secondary analysis. That exercise reveals whether your questionnaire or dataset actually contains the required items. It also helps distinguish a finding from interpretation. A chart may show a pattern; the discussion should explain what the pattern does and does not support.
- List every research question and its primary outcome.
- Check the legal and ethical route to each variable or document.
- Write definitions and collection timing in a data dictionary.
- Choose summaries, primary analysis and diagnostic checks.
- Specify missing-data reporting and decision rules for deviations.
- Sketch the results tables without inventing numbers.
- Have the plan reviewed before collection, and retain a dated version.
Planning sample size belongs here too: the target should reflect the primary outcome and analysis, not a generic questionnaire rule. If your question and hypothesis are not yet aligned, start with Labeeb’s research-question-versus-hypothesis guide.
Sources checked: NIH NCATS data analysis plan; Canadian Journal of Hospital Pharmacy research primer; NCBI research data stewardship; prospective protocol statistical principles; observational analysis-plan template. Follow the current programme handbook and supervisor requirements.
Scope the analysis before the dataset arrives
If you have a lawful dataset or an approved study plan but are uncertain whether the variables can answer your research questions, Labeeb can scope a methods and analysis review. We can help document cleaning, assumptions, outputs and an explanation you can defend. We do not fabricate data, choose results to suit a preferred conclusion or write a dissertation for submission. Share your questions, proposed variables, university guidance and deadline for a consultation.
Confirmatory Factor Analysis Results: Read Model Fit Without a Magic Cutoff
Rayyan Screening Workflow for a Systematic Review: Decisions and PRISMA Counts
Research Question vs Hypothesis: Choose and Align Your Dissertation
How to Justify Dissertation Sample Size: Power, Precision and Recruitment