Loading Which Statistical Test…
Quick guide: plain English, key facts and method
In plain English: Choosing a statistical test comes down to three decisions, like working through a flowchart.
Key facts
- Decision 1: is the researcher looking for a difference or a correlation?
- Decision 2: is the design related (repeated measures or matched pairs) or unrelated (independent groups)?
- Decision 3: are the data nominal, ordinal or interval?
- Parametric tests (the t-tests and Pearson’s r) need interval data and a normal distribution. The unrelated t-test also needs similar variances.
- Chi-squared needs frequencies (not percentages), with each participant counted in only one cell.
How to do it
- Underline the design and the way the DV is measured in the question.
- Make the three decisions in order and find the test in the grid.
- When you justify your choice, give all three reasons.
Memory aid: “Carrots Should Come Mashed With Swede Under Roast Potatoes” gives the grid in order: Chi-squared, Sign, Chi-squared (nominal); Mann–Whitney, Wilcoxon, Spearman (ordinal); Unrelated t, Related t, Pearson (interval). Within each level the order is unrelated, related, correlation.