Wilcoxon Test

Concept

---

Theoretically, the Wilcoxon Test is a non-parametric rank-based statistical test used to compare populations without assuming that the underlying data follow a normal distribution. Two principal forms are recognised: the Wilcoxon Signed-Rank Test for paired observations and the Wilcoxon Rank-Sum Test (equivalent to the Mann?Whitney U test) for independent samples. In survival analysis, the term "Wilcoxon test" commonly refers to the Breslow?Gehan Wilcoxon test, a weighted log-rank test that places greater emphasis on earlier events.

Mathematically, the Wilcoxon test replaces raw observations with their ranks and evaluates whether the distribution of ranks differs between groups. The test statistic is based on the sum of signed or grouped ranks and, for sufficiently large samples, is approximately normally distributed after standardisation. In survival analysis, the Breslow?Gehan Wilcoxon test weights each event by the number of individuals at risk, increasing sensitivity to early differences in survival.

In practice, Wilcoxon tests are widely applied when data are ordinal, skewed or violate normality assumptions. In health economics they are frequently used to compare costs, quality-of-life scores and other non-normally distributed outcomes, while the Breslow?Gehan Wilcoxon test is commonly used to compare survival curves when treatment effects occur predominantly during early follow-up.

---

Purpose

---

Used to compare two populations without assuming normally distributed data, or to compare survival distributions with greater emphasis on early events when using the Breslow?Gehan Wilcoxon variant.

---

Mathematical Formulae

Primary Formula

Signed-rank statistic:

W = ?R?

where R? denotes the signed ranks of the paired differences.

For large samples:

Z = (W ? E(W)) / �Var(W)

Supporting Formulae

Expected value:

E(W) = n(n + 1) / 4

Variance:

Var(W) = n(n + 1)(2n + 1) / 24

For the Breslow?Gehan Wilcoxon survival test:

?� = (?n?(O? ? E?))� / ?n?�V?

where n? is the number at risk at event time i.

Related Mathematical Methods

• Mann?Whitney U Test
• Wilcoxon Signed-Rank Test
• Breslow?Gehan Wilcoxon Test
• Weighted Log-Rank Test
• Rank-Based Statistical Methods
• Non-Parametric Hypothesis Testing

---

Example

---

A health economist compares EQ-5D utility scores before and after a rehabilitation programme in 40 patients. Because the paired differences are not normally distributed, a Wilcoxon Signed-Rank Test is performed.

The analysis produces:

W = 124

Z = ?2.61

p = 0.009

The results indicate a statistically significant improvement in health-related quality of life following treatment.

---

Excel Implementation

Function	Example Formula	Health Economics Application
RANK.AVG	=RANK.AVG(B2,$B$2:$B$41,1)	Rank observations or paired differences
ABS	=ABS(B2-C2)	Calculate absolute paired differences
SUM	=SUM(D2:D41)	Calculate the Wilcoxon rank sum
IF	=IF(B2-C2>0,1,-1)	Assign the sign of paired differences
NORM.S.DIST	=2*(1-NORM.S.DIST(ABS(Z),TRUE))	Approximate two-sided p-value for large samples

---

VBA (Optional)

---

A VBA procedure can automate Wilcoxon rank calculations, compute test statistics and generate summary reports for multiple clinical or health economic outcome comparisons.

---

Sources

• Wilcoxon F. Individual Comparisons by Ranking Methods. Biometrics Bulletin. 1945.
• Hollander M, Wolfe DA, Chicken E. Nonparametric Statistical Methods.
• Conover WJ. Practical Nonparametric Statistics.
• Collett D. Modelling Survival Data in Medical Research.
• Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation.