A public health researcher in Delhi wants to compare the mean systolic blood pressure between two independent groups of adult males: those with regular physical activity (n=45) and sedentary individuals (n=50). The data are normally distributed with unequal variances. Which statistical test is most appropriate to determine if there is a significant difference between the two groups?

## Test Selection for Comparing Two Independent Means ### Study Design Analysis - **Two independent groups** (not paired/matched) - **Continuous outcome** (systolic blood pressure in mmHg) - **Normal distribution** (parametric test applicable) - **Unequal variances** between groups - **Sample sizes** adequate (n₁=45, n₂=50) ### Why Welch's t-test (Unpaired t-test with unequal variances)? **Key Point:** Welch's t-test is the gold standard when comparing two independent groups with normally distributed data but unequal variances (heteroscedasticity). **High-Yield:** The classical Student's t-test assumes equal variances (homoscedasticity). When variances are unequal, Welch's correction adjusts the degrees of freedom using the Welch–Satterthwaite equation: $$df = \frac{\left(\frac{s_1^2}{n_1} + \frac{s_2^2}{n_2}\right)^2}{\frac{(s_1^2/n_1)^2}{n_1-1} + \frac{(s_2^2/n_2)^2}{n_2-1}}$$ This protects against Type I error inflation when variances differ. ### Comparison Table | Test | Use Case | Assumption | |------|----------|------------| | **Paired t-test** | Two dependent/matched groups | Normality, paired design | | **Student's t-test** | Two independent groups | Normality, **equal variances** | | **Welch's t-test** | Two independent groups | Normality, **unequal variances** | | **One-way ANOVA** | 3+ independent groups | Normality, equal variances | | **Chi-square test** | Categorical data, frequencies | Non-parametric | **Clinical Pearl:** In real-world research, unequal variances are common. Welch's t-test is more robust and is now the default recommendation in many statistical software packages (R, modern SPSS). ### Decision Algorithm ```mermaid flowchart TD A[Comparing two groups?]:::decision --> B{Independent or paired?}:::decision B -->|Paired| C[Paired t-test]:::action B -->|Independent| D{Data normally distributed?}:::decision D -->|No| E[Mann-Whitney U test]:::action D -->|Yes| F{Equal variances?}:::decision F -->|Yes| G[Student's t-test]:::action F -->|No| H[Welch's t-test]:::action ```