t-Test, ANOVA or Chi-Square: Which One Do I Need?
The t-test, ANOVA and chi-square test are three of the most widely used tests in student research. Each answers a different kind of question, and choosing the wrong one is a common mistake. In this blog, we'll explain when to use each, with simple examples.
The t-Test:
- Use a t-test to compare the means of two groups on a continuous outcome.
- Independent-samples t-test: two separate groups, for example comparing exam scores of students taught online and in person.
- Paired-samples t-test: the same people measured twice, for example stress scores before and after a training programme.
- Assumptions include a continuous outcome, approximate normality and, for the independent test, similar variances (SPSS reports Levene's test to help check this).
ANOVA:
- Use a one-way ANOVA to compare the means of three or more groups on a continuous outcome.
- Example: comparing job satisfaction across three departments.
- A significant ANOVA tells you that at least one group differs, but not which. Use post hoc tests, such as Tukey's test, to find out where the differences lie.
- Running several t-tests instead of one ANOVA increases the chance of a false positive result.
The Chi-Square Test:
- Use a chi-square test of independence to check whether two categorical variables are associated.
- Example: whether preferred study mode (online or in person) is related to gender.
- It compares observed frequencies with the frequencies you would expect if there were no association.
- It is less reliable when expected counts in cells are very small. In that case, Fisher's exact test may be more appropriate for small tables.
A Quick Way to Decide:
- Is your outcome continuous and your grouping variable categorical? Use a t-test for two groups or ANOVA for three or more.
- Are both variables categorical? Use a chi-square test.
- Are both variables continuous? You probably need correlation or regression instead.
Reporting the Results:
- t-test: report the t value, degrees of freedom, p value, group means and standard deviations.
- ANOVA: report F, degrees of freedom, p value and post hoc results.
- Chi-square: report the chi-square value, degrees of freedom, p value and the cross-tabulated counts or percentages.
- Where possible, also report an effect size, as recommended by many style guides.
The choice between a t-test, ANOVA and chi-square test depends on your variable types and the number of groups. Match the test to your question, check its assumptions, and report both the result and what it means in plain language.