Both Kendall’s tau and Spearman’s rank correlation tests measure the strength and direction of association between two ranked variables. Kendall’s tau evaluates the correspondence of orderings in pairs, while Spearman’s rank correlation assesses how well a monotonic function describes the relationship between variables. By ranking students' grades and applying these tests, one can test the null hypothesis that there is no correlation between midterm and final exam scores. If the p-value is less than the significance level (0.05), the null hypothesis is rejected, suggesting a significant positive correlation between midterm and final grades .

The Mann-Whitney U test determines whether there is a significant difference between the distributions of two independent samples—in this case, the breaking strengths from two processes. It does not assume normality, making it appropriate when data are not normally distributed. The test ranks all data across both groups collectively and relies on the assumption that data points in each group are independent and yield measures on at least an ordinal scale. The null hypothesis is that the distributions are identical, and thus, any observed rank difference is due to random chance .

The Mann-Whitney U test helps identify significant differences in breaking strength between products from manufacturing processes without assuming normal distribution. Detecting such differences is crucial for quality assurance and improvement. If the null hypothesis—that the processes produce items with equal strength—is rejected, it indicates that one process consistently yields stronger or weaker outputs. This insight can guide process optimization strategies, such as adjusting production parameters or further investigation into material properties and method consistency to enhance quality across all processed items .

The sign test only considers the direction (above or below) of discrepancies from the median, not their magnitude, making it less powerful than the Wilcoxon signed rank test, which uses both the direction and magnitude of the differences by ranking them. This makes the Wilcoxon test more sensitive to changes in the data, hence often more powerful when assumptions are met. Because of these differences, the Wilcoxon test is preferable when the ordinal nature of the data and the assumptions of symmetry around the median are justified, while the sign test is applicable when data do not meet such assumptions, offering a simpler and more robust alternative .

The Kruskal-Wallis test is preferable to ANOVA when the data does not meet the assumptions necessary for ANOVA, specifically the requirements for normal distribution and homogeneity of variances. It is used to test if there are statistically significant differences between the medians of three or more independent groups. In analyzing calculator operating times, this test would involve ranking operating times across all calculators and assessing if the groups have the same distribution of ranks, based on the null hypothesis that all group distributions are identical. Should these ranks deviate significantly, the null hypothesis is rejected at the chosen significance level, such as 0.01 in this instance .

The Wald-Wolfowitz runs test checks sequences for randomness by analyzing patterns of related values—in this case, defectives and non-defectives. A run is a sequence of identical elements, and the test assesses if the number of runs is consistent with randomness. Too few runs suggest clustering, while too many imply over-dispersion. Applied in quality control, it helps to determine if there is a random distribution of defective items over time. The null hypothesis assumes that the occurrences of defectives are random; deviations indicate potential systematic issues in the production process .

Non-parametric tests like the Mann-Whitney U test have several limitations. They generally have less power than parametric tests when the assumptions for parametric tests (normal distribution and equal variances) are met, which can make it harder to detect true effects. Non-parametric tests also assume that the data distributions of the groups are similar in shape; differences might lead to misleading conclusions about medians or central tendencies. These constraints can affect the interpretation of results, especially in distinguishing genuine findings from those influenced by deviations from test assumptions .

The sign test can be employed to evaluate whether the median of the waiting times is at most 20 minutes. It is a non-parametric test used when the assumptions for parametric tests cannot be met, mainly when the data does not follow a normal distribution. The key assumption of the sign test is that the data is continuous, and each observation pair can be categorized into either a positive or negative difference relative to the median. Any ties are ignored in the analysis. To perform this test, one counts the number of times the waiting time is above and below 20 minutes and checks if this distribution is significantly unbalanced at a specified significance level, such as 0.05 .

The Wilcoxon signed rank test provides a robust alternative to parametric methods, which can be significantly impacted by non-normal data distribution. This test does not require the assumption of normality and is suitable for small sample sizes, making it ideal for diet studies often limited by number of participants. It accounts for the direction and magnitude of weight changes, offering a nuanced analysis of the effect of a diet. This validates whether observed changes can be attributed to the diet itself rather than random variation, especially useful when the sample deviates from normality .

When choosing between the Wilcoxon signed rank test and the Mann-Whitney U test, the primary considerations are the data structure and the underlying hypotheses. The Wilcoxon signed rank test is used for paired data to assess differences in medians of two related groups or matched pairs. In contrast, the Mann-Whitney U test is suited for comparing two independent groups to detect differences in their central tendencies. Additionally, the choice may also depend on data scales; specifically, whether the data meet the assumptions of continuity and symmetry required for the Wilcoxon test, versus merely ordinal data required for the Mann-Whitney U test .