Short answer: work from your variables. What type is the outcome, what type is the predictor, and how many groups are you comparing? Answer those three and the test picks itself. The hardest skill in STAT115 is not running an analysis, it is deciding which one the scenario needs.
Why students find this hard
Because of how the course is practised, not how it is taught. If you do the t-test exercises in the t-test week, you never have to identify anything, the chapter already told you. Then the exam presents a scenario with no label and the identification step is one you have literally never rehearsed.
The three questions
1. What kind of outcome variable?
Continuous, like blood pressure or weight, or categorical, like diseased versus healthy. This single question eliminates most of the options immediately.
2. What kind of predictor or grouping variable?
Categorical groups, like treatment versus control, or another continuous variable, like age.
3. How many groups, and are they related?
Two groups or more than two. Independent groups, or paired measurements on the same people before and after.
The rough map
- Continuous outcome, two independent groups: two-sample t-test.
- Continuous outcome, paired measurements: paired t-test.
- Continuous outcome, three or more groups: ANOVA.
- Continuous outcome, continuous predictor: correlation or linear regression.
- Categorical outcome and categorical predictor: chi-square test.
- Categorical outcome with several predictors: you are into the models-for-binomial-data territory the course covers later.
Keep this map somewhere you can rebuild it from memory, because being able to reconstruct it under pressure is the exam skill.
Train identification, not computation
Practise in mixed sets. Read a scenario, state the outcome type, the predictor type and the number of groups, then name the test, all before touching a calculation. Interleaved practice is exactly what builds this discrimination, and it beats blocked practice on later tests despite feeling harder (Rohrer & Taylor, 2007).
You can do a lot of this without computing anything at all. Twenty scenarios where you only name the test is a better use of thirty minutes than two full worked analyses.
Watch the assumptions
Choosing the right family is most of the mark, but STAT115 also cares whether the assumptions hold: roughly normal distributions for the t-tests, independence of observations, adequate expected counts for chi-square. A question that looks like a t-test but violates an assumption is a favourite way to separate students.
Interpretation is the other half
Running the test is the easy part, especially with R doing the arithmetic. What the exam actually rewards is saying what the result means in context: what the p-value does and does not tell you, what the confidence interval implies, and whether the conclusion is justified. Practise writing that sentence, not just producing the number.
You can practise scenario-based STAT115 questions free on Cutline.
References
- Rohrer, D., & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498.
