Answer each of the following questions in about 150 words :
Explain parametric test and non-parametric test. What are the applications of chi-square (χ²) test? Give one example.
10 marks · answer in 150 words · directive: Explain
The core of a good answer
Parametric tests assume a known population distribution (usually normal) and test hypotheses about its parameters, while non-parametric tests make few or no distributional assumptions; the chi-square test, a non-parametric test on frequencies, is used for goodness of fit, independence of attributes and homogeneity.
An outline written by upsc.help, not an official UPSC answer; every fact checked by a separate review. How we work
What the question asks
The question is split 3+7: a short, clear contrast between parametric and non-parametric tests, then the bulk on the uses of the chi-square test with one worked or described example. A small numerical illustration of a test of independence earns the most credit in 150 words.
Answer outline
Introduction
A statistical test checks a hypothesis about a population using sample data; tests are classed as parametric or non-parametric according to the assumptions they make about the population.
Parametric tests
- Assume the form of the population distribution (normally normal) and test hypotheses about its parameters such as mean or variance.
- Rely on assumptions such as normality, independent samples and, for some tests, equal variances.
- Examples: Z test, t test (one sample, paired, two sample), F test and ANOVA.
- More powerful than non-parametric tests when their assumptions hold.
Non-parametric tests
- Distribution-free: make few or no assumptions about the shape of the population distribution.
- Work on categorical or ordinal data, ranks or signs, and suit skewed data.
- Examples: chi-square test, sign test, Wilcoxon signed-rank test, Mann-Whitney U test, Kruskal-Wallis H test, Spearman's rank correlation.
- Simpler to apply but less efficient when parametric assumptions are actually met.
Applications of the chi-square test
- Goodness of fit: whether observed frequencies fit a theoretical distribution (uniform, binomial, Poisson, normal); df = number of classes minus 1 minus parameters estimated.
- Test of independence of attributes in an r × c contingency table, e.g. whether brand preference depends on income group; df = (r - 1)(c - 1).
- Test of homogeneity: whether several populations share the same distribution of a categorical variable.
- Statistic: χ² = Σ (O - E)² / E, compared with the table value at the chosen level of significance.
Example: test of independence
- 200 customers: 100 saw an advertisement (60 bought, 40 did not); 100 did not see it (30 bought, 70 did not).
- H0: buying is independent of seeing the advertisement. Expected frequency for each cell = row total × column total / 200, giving 45 and 55 in each row.
- χ² = (60-45)²/45 + (40-55)²/55 + (30-45)²/45 + (70-55)²/55 = 5 + 4.09 + 5 + 4.09 = 18.18 (approximately).
- df = (2-1)(2-1) = 1; table value at 5 per cent = 3.841. Since 18.18 > 3.841, H0 is rejected: the advertisement and purchase are associated.
Conclusion
Parametric tests are preferred when their assumptions hold; the chi-square test is the main non-parametric tool for categorical data in market research, quality control and HR studies.
Examples and data to cite
- Karl Pearson first investigated the properties of the chi-square test in 1900.
- Typical management uses: brand preference by region, defect rates across shifts, attrition by department, survey response patterns.
Facts to remember
- χ² = Σ (O - E)² / E; expected frequency in a contingency table = (row total × column total) / grand total.
- Degrees of freedom: (r - 1)(c - 1) for independence and homogeneity; k - 1 - m for goodness of fit with m estimated parameters.
- Common rule: expected frequencies of 5 or more in all cells of a 2 × 2 table and in at least 80 per cent of cells of larger tables; Yates' correction (reducing each |O - E| by 0.5) is used with 1 df when counts are small.
- Critical value of χ² for 1 df at 5 per cent level: 3.841.
- Chi-square values are always non-negative and the test is usually one-tailed (right tail).