Write the analysis plan before you collect data

An analysis plan is a short document that states, before the data exist, how they will be analysed. It protects you from two risks: collecting data that cannot answer your question, and making analytic choices after seeing the results.

What the plan contains

• The research questions and hypotheses, numbered.
• The variables: how each is measured, its scale and its role in the analysis.
• The primary outcome and the primary analysis for each hypothesis.
• The sample size and how it was determined.
• How missing data, outliers and protocol deviations will be handled.
• Assumption checks, and what you will do if they fail.
• Secondary and exploratory analyses, labelled as such.
• The software and its version.

Why it helps

Writing the plan exposes gaps while they are still cheap to fix. You may discover that a key variable is missing from the questionnaire, or that the design gives you no comparison group. It also shortens the methods section later, because most of it is already written.

What it does not do

A plan does not forbid changes. Research meets surprises. When you depart from the plan, record what changed and why, and report it. Readers trust an analysis more when the planned and unplanned parts are clearly separated.

Getting started

One or two pages are enough for most theses. Share the plan with your supervisor or a statistician before data collection and keep a dated copy. In clinical research, and in a growing number of other fields, the plan can be registered publicly.