Which of the following is true for non-parametric tests?

Explanation

## Correct Answer: D. It is used for skewed distributions Non-parametric tests are statistical methods that do not assume the data follows a normal (Gaussian) distribution. The key discriminating feature is their applicability to **skewed distributions**. Unlike parametric tests (t-test, ANOVA) which require normality assumptions, non-parametric tests are distribution-free and work with ordinal, ranked, or heavily skewed continuous data. In Indian clinical practice, many real-world datasets—such as hospital admission rates, disease severity scores, or patient outcome measures in resource-limited settings—are often skewed rather than normally distributed. Non-parametric tests (Mann-Whitney U, Kruskal-Wallis, Spearman's rank correlation) convert raw data to ranks, making them robust against outliers and skewness. This is why they are preferred in epidemiological surveys and public health studies where data distribution is unknown or clearly non-normal. The test statistic depends on ranks rather than actual values, eliminating the need for normality assumptions. ## Why the other options are wrong **A. It cannot be used for small sample sizes** — This is the opposite of reality. Non-parametric tests are **particularly useful for small sample sizes** because they do not require the large-sample normality assumption that parametric tests depend on. When n < 30, non-parametric tests are often more appropriate. This is a classic NBE trap—students confuse the advantage of non-parametric tests with a limitation. **B. It involves the assumption that the data has a normal distribution** — This is the defining characteristic of **parametric** tests, not non-parametric tests. Non-parametric tests are explicitly designed to avoid normality assumptions. This is a direct reversal trap—NBE tests whether students know the fundamental distinction between parametric and non-parametric approaches. **C. ANOVA is an example of a non-parametric test** — ANOVA (Analysis of Variance) is a **parametric test** that assumes normal distribution and homogeneity of variance. The non-parametric equivalent of ANOVA is the **Kruskal-Wallis test**. This option confuses students who know ANOVA is used for comparing multiple groups but don't recall its parametric nature. ## High-Yield Facts - **Non-parametric tests** are distribution-free and do not assume normality; they work with ranked or ordinal data. - **Skewed distributions** (common in Indian epidemiological data) are the primary indication for non-parametric tests over parametric alternatives. - **Mann-Whitney U test** is the non-parametric alternative to independent t-test; **Kruskal-Wallis** replaces ANOVA. - **Small sample sizes** (n < 30) are an advantage for non-parametric tests, not a limitation. - **Spearman's rank correlation** is used instead of Pearson's r when data is ordinal or non-normally distributed. ## Mnemonics **RANK = Non-Parametric** Non-parametric tests convert data to RANKS, making them work with any distribution shape. Parametric tests use RAW values and need NORMAL distribution. **SKEW → Non-Parametric** When your data is SKEWED (asymmetric, outliers present), reach for non-parametric tests. Parametric tests assume bell curve (symmetric). ## NBE Trap NBE pairs "small sample sizes" with non-parametric tests to lure students into choosing option A—students often confuse the *advantage* (works well with small n) with a *limitation*. The correct understanding is that non-parametric tests are **especially valuable** when n is small and normality cannot be assumed. _Reference: Park's Textbook of Preventive and Social Medicine, Ch. 10 (Biostatistics); Guyton & Hall Medical Physiology, Statistical Methods section_