MKTG90011 · Marketing Research
Testing Differences
This is the heart of the subject: from a clean dataset to the right test. The anchor skill is the which-test framework — pick the test from three facts: the scale of the DV and IV, the number of groups, and whether groups are paired or independent. Before any test you clean the data and describe it with the right statistic for each scale. Then, for comparing means on a metric outcome, the t-test family handles two situations and ANOVA handles more: a one-sample t-test compares one group's mean to a benchmark; an independent-samples t-test compares two separate groups; a paired-samples t-test compares the same subjects twice (before/after). When you have 3+ groups, a single t-test is illegal — you use one-way ANOVA, read its F statistic and Sig., and if significant run a post-hoc test (Tukey) to locate which groups differ. The exam hands you a hypothesis and SPSS output and asks you to name the test, justify it by structure, and report the decision in one APA sentence; the project mandates two different t-tests (H2, H3) and a one-way ANOVA, so all three flavours must be reflex.
What this chapter covers
- 017.1 Data preparation — clean before you compute
- 02Descriptive statistics matched to the scale type
- 03The logic of hypothesis testing (H₀ / H₁, p, α)
- 04★ The which-test decision framework
- 05The t-statistic = mean difference / standard error
- 06One-sample, independent-samples and paired-samples t-tests
- 07One-way ANOVA for 3+ groups — F, Sig. and post-hoc
Worked example: choose the difference test and report it
- +1(a) Pick the test. One metric outcome, two measurements on the same people (before/after) → paired-samples t-test (independent would be wrong — the groups aren't separate).
- +1(b) Hypotheses. H₀: the mean before equals the mean after (mean difference = 0); H₁: the means differ (mean difference ≠ 0).
- +1Read the statistic. t = 3.4 with df = 79; the two-sided Sig. = .001, well below α = .05.
- +1(c) Decide. p < .05 → reject H₀: satisfaction changed significantly after the redesign.
- +1Report (APA). “Satisfaction rose significantly after the redesign, t(79) = 3.40, p = .001” — state direction, the statistic, df and p.
Key terms
- Which-test framework
- The decision rule that selects a statistical test from the scale of each variable, the number of groups, and whether groups are paired or independent. It is the single most-examined skill in MKTG90011.
- t-test
- A test comparing means on a metric DV, where t = (mean difference) / (standard error of that difference). A large |t| and small p mean the means are unlikely to be equal, so you reject H₀.
- Independent vs paired samples
- Independent-samples t compares two separate groups (e.g. men vs women); paired-samples t compares the same subjects measured twice (before/after, or matched pairs). The data structure, not the topic, decides which.
- One-way ANOVA
- The test for comparing the means of a metric DV across three or more independent groups. It returns an F statistic and a Sig.; a significant result triggers a post-hoc test to find which groups differ.
- Null and alternative hypotheses
- H₀ states no effect (means equal); H₁ states an effect. You reject H₀ when the test's Sig. (p) falls below the significance level α (usually 0.05).
Testing Differences FAQ
How do I choose between an independent and a paired t-test?
By the data structure. If the two sets of scores come from separate people (men vs women, treatment vs control), use the independent-samples t-test. If they come from the same people measured twice (before/after) or matched pairs, use the paired-samples t-test. The topic is irrelevant; only who the scores belong to matters.
Why can't I just run several t-tests instead of ANOVA?
Because each t-test carries a 5% false-positive risk, and running many inflates the overall chance of a spurious “significant” result. One-way ANOVA tests all 3+ group means at once with a single F test, controlling that error; you only drill into specific pairs afterwards with a post-hoc test.
What do I actually read off the SPSS output?
The statistic (t or F), its degrees of freedom, and the Sig. (p-value). If Sig. < .05 you reject the null. Then write the one-line APA report — e.g. t(79) = 3.40, p = .001 — stating the direction of the effect. For ANOVA, follow a significant F with a post-hoc (Tukey) to say which groups differ.
What does the p-value tell me (and not tell me)?
It is the probability of a result this extreme if H₀ were true; below α (0.05) you call the result statistically significant. It does not measure the size or practical importance of the effect — a tiny, trivial difference can be significant in a large sample, so always read the means alongside the p.
Exam move
This chapter is where most exam marks live, so make the which-test decision tree automatic: scale of each variable → number of groups → paired or independent. Learn the three t-test flavours by their structure (one group vs benchmark, two separate groups, same group twice) rather than by topic, and know that 3+ groups forces one-way ANOVA. For every test, practise the read-and-report drill — statistic, df, Sig., decision, one APA sentence — because the exam grades the report, not the computation. The project's H2, H3 and H4 use exactly these tests, so the same preparation serves both.
