Why this chapter matters
Many real questions involve group comparisons or categorical relationships rather than a single mean or proportion.
What you will learn
- Apply two-sample confidence intervals and tests for means and proportions.
- Use chi-square tests for independence and goodness of fit with expected-count checks.
- Interpret statistically significant differences with careful context language.
Understand the core ideas
Many applied questions compare groups or evaluate categorical patterns, so one-parameter tools are not enough. Two-sample inference addresses differences between means or proportions, with procedure choice based on variable type and whether groups are independent or paired. For categorical tables, chi-square methods compare observed counts with expected counts under a null model of independence or a claimed distribution. Expected-count checks are important because very small expected values can make approximation inaccurate.
Interpretation should focus on what the test does and does not show. A significant two-sample result suggests a population difference is plausible, but causation still depends on design and confounding control. A significant chi-square independence test indicates association pattern in counts, but it does not by itself identify which cells drive the signal or quantify practical importance. Follow-up summaries and effect-size context are needed for responsible reporting. Teams should document assumptions and expected-count checks so later readers can verify method choices quickly.
Key terms
- Two-sample t procedure
- Method for comparing population means from two groups using sample data.
- Difference in proportions
- Contrast between two group proportions used for estimation or testing.
- Chi-square statistic
- Sum of squared observed-expected differences scaled by expected counts.
- Expected count
- Count predicted by the null model for a table cell or category.
Two-proportion comparison
Group A has 54 passes out of 80. Group B has 40 passes out of 80.
- Compute sample proportions: p1-hat = 0.675 and p2-hat = 0.500.
- Compute observed difference: 0.175.
- Under H0 equal proportions, compute pooled p-hat = 94/160 = 0.5875 and standard error approximately 0.0778.
- Compute z = 0.175 / 0.0778 approximately 2.25 and two-sided p-value about 0.024.
A common misconception
Claim: A significant chi-square test proves a strong and important relationship.
Correction: Significance indicates evidence of association under the model; strength and practical impact require additional measures and context.
Lessons in this chapter
- Two-sample means and proportionsChoose and run two-sample procedures based on variable type and design.
- Chi-square for independenceEvaluate whether two categorical variables are associated.
- Chi-square goodness of fitCompare observed counts to a claimed categorical distribution.
- Practical interpretation of group differencesReport direction, size, and uncertainty of detected differences.
Study task
Chapter checkpoint
What is the null hypothesis in a chi-square test of independence?
The two categorical variables are independent in the population.