Non-Parametric Methods

Introduction: Non-Parametric Methods

Non-parametric methods are statistical techniques that do not rely on assumptions about data following a normal distribution. They are flexible, robust tools for analysing data that is skewed, ordinal, or includes outliers. These methods provide reliable insights even when traditional assumptions about normality and variance are violated.

Background

While parametric tests (such as the t-test or ANOVA) are powerful, they depend on assumptions like normality and equal variances. When those conditions are not met, results can be misleading.
Non-parametric methods were developed to address this limitation. They focus on the rank or order of values rather than their precise magnitude, making them more suitable for ordinal or non-normally distributed data.

Commonly used non-parametric tests include the Sign Test, Wilcoxon Signed-Rank Test, Mann–Whitney U Test, and Kruskal–Wallis Test.

Key Elements / Features

• Distribution-free: Do not assume data follows a specific distribution.
• Robust: Less affected by outliers or skewed data.
• Rank-based: Use ranks instead of raw scores for comparison.
• Broad scope: Cover a range of tests for one-sample, two-sample, and multi-sample problems.

Example formula (Wilcoxon Signed-Rank Test statistic):

\(
W = \sum R_i
\)

Where:

• W = sum of the signed ranks
• Ri = rank of each observation’s absolute difference

Applications / Examples

• Customer surveys: Analysing satisfaction ratings on ordinal scales (e.g., 1–5).
• Healthcare: Comparing patient outcomes in small or non-normally distributed samples.
• Business and service processes: Evaluating feedback or response times with outliers.
• Quality improvement: Assessing process changes when data is non-continuous or categorical.

Example:
A Lean team compares customer satisfaction before and after a process change using a Wilcoxon Signed-Rank Test instead of a t-test, since the data are ordinal (Likert scale).

Relevance / Impact

Non-parametric methods extend statistical analysis to real-world data that rarely fits ideal assumptions. They enhance decision-making in Lean Six Sigma by providing valid results from small, irregular, or skewed datasets — ensuring improvement decisions are based on sound, data-driven evidence.

See also

• 1-Sample Sign Test
• Wilcoxon Signed-Rank Test
• Mann–Whitney U Test
• Kruskal–Wallis Test