A common difficulty in statistics assignments is choosing a method that actually matches the research question and data. Rather than selecting a test from a memorised list, work from the outcome variable, the comparison being made, the study design and the assumptions. This guide provides a starting framework for introductory coursework; your instructor's methods and requirements take precedence.
Define the research question and variables
Write the question in one sentence and identify the outcome variable and any predictor or grouping variable. Determine whether each variable is numerical, categorical, ordinal, paired or repeated over time. The measurement scale and data structure often determine which methods are plausible.
For example, comparing the average exam score of two independent classes is different from comparing the same students' scores before and after a course. The first uses independent groups; the second uses paired measurements.
Understand common test families
A t-test is often used for a question about means under specified conditions. ANOVA is a common option for comparing means across more than two groups. A chi-square test can assess relationships between categorical variables or compare observed counts with expected counts. Correlation describes association between variables, while regression models an outcome in relation to one or more predictors.
These are not interchangeable labels. Check whether observations are independent, whether the design is paired, what distributional assumptions apply, and whether the sample and outcome support the intended method. Some data call for non-parametric or specialised methods instead.
Use a decision checklist
Begin with the outcome and the comparison. If the outcome is categorical, a count-based method may be appropriate; if the outcome is numerical and the question concerns group means, a mean-comparison method may be appropriate. If the task studies a relationship between numerical variables, correlation or regression may be relevant.
Treat the following as a first-pass study aid rather than a substitute for checking test assumptions, sample design or the exact problem statement.
- Two independent groups and a numerical outcome: consider an independent-samples t-test if assumptions are suitable.
- The same participants measured twice: consider a paired t-test if assumptions are suitable.
- Three or more independent group means: consider ANOVA under its assumptions.
- Two categorical variables in a contingency table: consider a chi-square test when count assumptions are met.
- Association between numerical variables: consider correlation or regression based on the question.
- A prediction question: define the outcome, predictors and model assumptions before fitting a model.
Report the result correctly
Your write-up should describe the method, why it fits the research question, the relevant test statistic and p-value, and an effect estimate or confidence interval where required. Explain the conclusion in context instead of writing only 'significant' or 'not significant'. A small p-value does not by itself show that an effect is large or practically important.
Avoid saying that you proved the null hypothesis when a result is not statistically significant. State the limits of the data, test assumptions and study design.
Common mistakes in statistics coursework
Students sometimes choose a test based only on the number of groups, ignore whether data are paired, treat ordinal ratings as automatically continuous, or report a p-value without the question's context. Another error is selecting a test after seeing the result rather than according to a planned analysis.
Write down your reasoning before calculating. If the design or assumptions are unclear, consult the assignment brief, course notes or instructor rather than guessing.
- Match the test to the research question and data type.
- Check independence, pairing, sample structure and assumptions.
- Distinguish statistical significance from practical importance.
- Report effect size or uncertainty when the assignment asks for it.
- Do not overstate conclusions or imply causation from association alone.
Understanding the concept is the first step.
Explore related resources or learn about subject-focused tutoring designed to help you understand the material.