The test of significance done for two or more proportions?

Explanation

## Correct Answer: A. Chi- square test The **chi-square test** is the gold-standard test of significance for comparing two or more proportions (categorical data). It tests the null hypothesis that observed frequencies match expected frequencies in categorical variables. When you have categorical outcomes (yes/no, diseased/non-diseased, exposed/unexposed) across two or more groups, chi-square is the appropriate parametric test. For example, comparing disease prevalence across three different Indian states, or comparing vaccination coverage rates across multiple districts—these are proportion comparisons requiring chi-square. The test statistic is calculated as $\chi^2 = \sum \frac{(O-E)^2}{E}$, where O = observed frequency and E = expected frequency. Chi-square assumes independence of observations and adequate cell frequencies (typically ≥5 in each cell). It is widely used in epidemiological surveys, cross-sectional studies, and case-control studies in Indian public health research (RNTCP surveillance, NFHS data analysis). ## Why the other options are wrong **B. ANOVA test** — ANOVA (Analysis of Variance) is used to compare **means** of continuous data across three or more groups, not proportions. It tests differences in quantitative variables (e.g., hemoglobin levels, blood pressure across age groups). Proportions are categorical, not continuous—ANOVA is inappropriate here. This is a common trap for students who confuse 'multiple groups' with 'multiple comparisons.' **C. Z test** — The Z test compares proportions between **exactly two groups only** (e.g., disease rate in vaccinated vs. unvaccinated). When you have three or more proportions to compare simultaneously, Z test cannot handle the multiple comparisons problem and lacks the framework to test overall association. Chi-square extends this logic to multiple groups. **D. Student's test** — Student's t-test (parametric) compares **means** of continuous variables between two groups. It is for quantitative data (e.g., comparing average weight, height, or lab values). Proportions are categorical data, not continuous—t-test is fundamentally wrong for this scenario. This option confuses categorical vs. continuous data types. ## High-Yield Facts - **Chi-square test** is the test of choice for comparing two or more **proportions** (categorical data). - Chi-square assumes **independence of observations** and minimum expected frequency of ≥5 per cell; if violated, use Fisher's exact test or Yates' correction. - **Z test** compares proportions in exactly **two groups**; chi-square extends to three or more groups. - **ANOVA** tests differences in **means** (continuous data), not proportions; **t-test** also requires continuous outcome variables. - Chi-square is widely used in Indian epidemiological surveys (NFHS, RNTCP) to compare disease prevalence, vaccination coverage, and risk factor prevalence across regions or populations. ## Mnemonics **CAP for Categorical Analysis** **C**ategorical data → **Chi-square**; **A**verage/continuous → **ANOVA**; **P**roportion (two groups) → **Z-test**. Use when deciding which test to apply in epidemiological data. **2+ Groups, Proportions = Chi-square** If you see 'two or more' + 'proportions/categorical,' immediately think chi-square. It's the workhorse of public health surveillance and cross-sectional studies. ## NBE Trap NBE pairs "two or more" with ANOVA to trap students who focus on the phrase "multiple groups" without distinguishing between continuous (means) and categorical (proportions) data types. The key discriminator is the **nature of the outcome variable**, not just the number of groups. ## Clinical Pearl In Indian NFHS surveys and RNTCP tuberculosis surveillance, chi-square is routinely used to compare disease prevalence or vaccination coverage across states, districts, or socioeconomic groups. A district TB officer comparing cure rates across three treatment centers would use chi-square—not ANOVA or t-test—because the outcome is categorical (cured/not cured). _Reference: Park's Textbook of Preventive and Social Medicine, Ch. 10 (Biostatistics); Mahajan's Methods in Biostatistics, Ch. 8 (Tests of Significance)_