Track 14 · Practical Data Analysis · Lesson 9 of 11
Selecting a method
Choose a statistical method by matching the question, the type of outcome and the structure of the data — and recognise when a tempting method does not fit.
30 minIntermediate8 graded exercisesMembers
A team lead sends you a spreadsheet of satisfaction scores taken before and after a new rota and asks whether things improved. Run the wrong test — treating the before and after scores as two unrelated groups — and a real improvement can look like noise. Choosing the method is a decision you should be able to justify in one sentence.
There is no need to memorise a giant flowchart. Almost every method choice follows from three questions:
- What is the question? Describe, compare groups, relate two variables, predict an outcome, or forecast a series over time.
- What type is the outcome? Numeric (a measurement or amount), categorical (a label; binary when there are two), ordinal (ordered labels such as a 1–5 rating), a count, or a value recorded at regular times.
- How are the data structured? How many groups? Are they independent (different units in each group) or paired (two measurements on the same unit)? How large are the samples, and how skewed?
Included with a Formula Quest membership
It has 8 graded exercises with feedback, hints and worked solutions. After it you will be able to:
- Choose a method from the question, the outcome type and the data structure
- Tell paired designs from independent groups, and use paired methods when measurements are linked
- Pick rank-based alternatives for ordinal or strongly skewed outcomes, and chi-square only for counts
- Run the chosen method in the lab and justify the choice in a sentence