Non-parametric Tests: Mann-Whitney U and Wilcoxon Signed-Rank Tests
Statistics Student Guide
Non-parametric Tests: Mann-Whitney U and Wilcoxon Signed-Rank Tests
Everything you need to understand non-parametric tests — when to use them, how to perform them step-by-step, how to interpret results, and how to report with proper effect sizes the way examiners and journal reviewers expect.
The Foundation
What Are Non-parametric Tests?
Non-parametric tests sit at the boundary of what makes statistics genuinely useful in the real world. Most classic tests — the t-test, ANOVA, Pearson’s correlation — are parametric: they assume your data follows a specific distribution, almost always normal, and they estimate parameters like the mean and standard deviation. When those assumptions hold, parametric tests are powerful. But data is rarely so cooperative.
When your data is ordinal, skewed, bounded, or collected from small samples where you can’t verify normality, parametric tests can produce misleading results. Non-parametric tests are the solution. They make no assumption about the underlying distribution. Instead, they work by ranking data — converting raw values into their order positions — and performing inference on those ranks. This makes them genuinely distribution-free. Statistics assignment help for students frequently begins with the question of which test to choose, and understanding parametric versus non-parametric is the foundational decision in any analysis.
~95%
Statistical efficiency of Mann-Whitney U vs. t-test when normality actually holds
>100%
Relative efficiency of non-parametric tests when data is non-normal — they outperform parametric tests
0
Normality assumptions required — the defining feature of non-parametric inference
Why Do Non-parametric Tests Work Without Normality?
The logic is elegant. By converting data to ranks, you strip away information about the raw scale — and with it, the dependence on distribution shape. The number 47 and the number 1,000,000 both become their rank positions. What matters is which observation is larger, not by how much. This rank transformation stabilizes the distribution of the test statistic under the null hypothesis, which is what lets us calculate valid p-values without assuming normality.
What Is the Difference Between Parametric and Non-parametric Tests?
Parametric tests estimate distributional parameters (mean, variance) and test hypotheses about those parameters. Non-parametric tests test hypotheses about ranks or medians without estimating distribution parameters. The key trade-off: parametric tests are more statistically powerful when their assumptions are met — meaning they are more likely to detect a real effect. Non-parametric tests sacrifice a small amount of power in exchange for robustness. When assumptions are violated, non-parametric tests actually become more powerful than their misbehaving parametric counterparts.
“Non-parametric tests are not the second-best option. For ordinal data and small non-normal samples, they are the correct option. Choosing a t-test on Likert scale data is not safer — it is wrong.” — A perspective widely taught in research methods courses at LSE, UCL, and the University of Michigan.
Decision Framework
When to Use Non-parametric Tests: The Decision Framework
Choosing between parametric and non-parametric tests is one of the most common decision points in applied statistics — and one of the most frequently botched. The answer depends on your data, your sample size, and your research design. Here’s the framework that works.
Conditions That Call for Non-parametric Tests
You should use a non-parametric test when at least one of these conditions applies to your data or design. First: your dependent variable is ordinal. Likert scales (Strongly Disagree to Strongly Agree), pain ratings (1–10), class rankings — these are ordinal. They have order but no guaranteed equal intervals between values. Treating them as continuous for a t-test is a debated practice; treating them with non-parametric tests is unambiguously defensible.
Second: your data is clearly non-normal and your sample is small. With large samples (n ≥ 30), the Central Limit Theorem kicks in and the sampling distribution of the mean becomes approximately normal regardless of the raw data distribution — so t-tests become robust. With small samples, you cannot rely on the CLT. If a Shapiro-Wilk normality test or Q-Q plot reveals significant departures from normality with n < 20, non-parametric tests are safer. Third: your data contains extreme outliers that distort means. Ranks are completely insensitive to outlier magnitude — the largest value always gets rank N regardless of whether it’s 100 or 1,000,000.
What Tests Do You Choose for Which Design?
Two Independent Groups
Parametric: Independent Samples t-test
Non-parametric: Mann-Whitney U Test (Wilcoxon Rank-Sum Test)
Example: Comparing exam scores between a lecture group and a flipped-classroom group where students are different people.
Two Related/Paired Groups
Parametric: Paired Samples t-test
Non-parametric: Wilcoxon Signed-Rank Test
Example: Comparing pain scores before and after treatment in the same patients, or comparing matched pairs of students.
If you have three or more independent groups, the non-parametric alternative to one-way ANOVA is the Kruskal-Wallis test. For three or more related groups, use the Friedman test. This article focuses specifically on the two-group case — Mann-Whitney U and Wilcoxon Signed-Rank — because they are the most commonly tested in university statistics courses and appear most frequently in published research.
Common mistake: Students confuse the Mann-Whitney U test and the Wilcoxon Signed-Rank test because both involve ranks. The distinction is fundamental — independent groups (different people) → Mann-Whitney U; paired/related observations (same people or matched pairs) → Wilcoxon Signed-Rank. Using the wrong test invalidates your analysis completely.
Struggling With Non-parametric Test Assignments?
Our statistics experts can help you choose the right test, run the analysis, and interpret results — with step-by-step explanations and fast turnaround.
Get Statistics Help Now Log InTest One
The Mann-Whitney U Test: Complete Guide
The Mann-Whitney U test — formally proposed by Henry Mann and Donald Whitney at Ohio State University in 1947, building on earlier work by Frank Wilcoxon — is the most widely used non-parametric test for comparing two independent groups. It is also known as the Wilcoxon rank-sum test, a name that reflects its computational basis. Despite the different names, both refer to the same procedure.
What Does the Mann-Whitney U Test Actually Test?
The formal null hypothesis of the Mann-Whitney U test is that the two populations have the same distribution. In practice — when the distributions have similar shapes — this is equivalent to testing whether the two group medians are equal. More precisely, it tests for stochastic dominance: the probability that a randomly selected observation from Group 1 exceeds a randomly selected observation from Group 2 equals 0.5.
H₀: P(X > Y) = 0.5 (both groups are equally likely to have larger values)
H₁: P(X > Y) ≠ 0.5 (two-tailed) or P(X > Y) > 0.5 or P(X > Y) < 0.5 (one-tailed)
Where X is a random observation from Group 1
and Y is a random observation from Group 2.
Assumptions of the Mann-Whitney U Test
- Independence: Observations within and between both groups must be independent.
- Ordinal or continuous measurement: The dependent variable must be at least ordinal.
- Similar distribution shape (for median comparison): If your goal is to compare medians, the two groups should have similarly shaped distributions.
How to Perform the Mann-Whitney U Test: Step-by-Step
Group A (Online): 72, 65, 88, 55, 79
Group B (In-person): 81, 70, 95, 63, 85, 77
n₁ = 5 (Group A), n₂ = 6 (Group B), N = 11 total
1
State the Hypotheses
H₀: The satisfaction score distributions are the same for online and in-person tutoring. H₁: The distributions differ (two-tailed). α = 0.05.
2
Combine and Rank All Observations
Score: 55(A), 63(B), 65(A), 70(B), 72(A), 77(B), 79(A), 81(B), 85(B), 88(A), 95(B)
Rank: 1 2 3 4 5 6 7 8 9 10 11
3
Calculate Rank Sums for Each Group
Group A ranks: 1 + 3 + 5 + 7 + 10 = 26 → R₁ = 26
Group B ranks: 2 + 4 + 6 + 8 + 9 + 11 = 40 → R₂ = 40
Verification: R₁ + R₂ = 26 + 40 = 66 = N(N+1)/2 = 11×12/2 = 66 ✓
4
Calculate U Statistics
U₁ = n₁·n₂ + n₁(n₁+1)/2 – R₁ = 5×6 + 5×6/2 – 26 = 30 + 15 – 26 = 19
U₂ = n₁·n₂ + n₂(n₂+1)/2 – R₂ = 30 + 6×7/2 – 40 = 30 + 21 – 40 = 11
Check: U₁ + U₂ = 19 + 11 = 30 = n₁×n₂ ✓
U = min(U₁, U₂) = min(19, 11) = 11
5
Find the p-value and Report Results
For n₁ = 5, n₂ = 6 at α = 0.05 two-tailed, critical value = 3. Since U = 11 > 3, we fail to reject H₀ (p > 0.05). No statistically significant difference found.
For larger samples, convert to z:
μ_U = n₁·n₂/2 = 15
σ_U = √(n₁·n₂(n₁+n₂+1)/12) = √30 ≈ 5.48
z = (U – μ_U)/σ_U = (11 – 15)/5.48 ≈ -0.73
Effect size: r = |z|/√N = 0.73/√11 ≈ 0.22 (small-medium)
Report: “A Mann-Whitney U test indicated no statistically significant difference in satisfaction scores between online (Mdn = 72) and in-person (Mdn = 79) tutoring groups, U = 11, z = -0.73, p > .05, r = .22.”
Key insight: The U statistic directly counts the number of times an observation from Group 1 “wins” against an observation from Group 2. U₁ = 19 means that for 19 out of 30 possible Group A vs. Group B comparisons, the Group A score was higher. This intuitive interpretation makes the Mann-Whitney U test conceptually powerful and practically meaningful.
Test Two
The Wilcoxon Signed-Rank Test: Complete Guide
The Wilcoxon Signed-Rank test was introduced by Frank Wilcoxon in a landmark 1945 paper, making it one of the oldest non-parametric tests in the statistical literature. It handles paired data — situations where each observation in one group has a corresponding observation in the other. Pre-test/post-test designs, before-and-after intervention studies, and matched-pairs experiments all call for this test.
What Is the Wilcoxon Signed-Rank Test Testing?
The test goes beyond the simple sign test by incorporating both the direction (sign) and magnitude (rank) of differences between paired observations. The null hypothesis is that the median of the paired differences is zero.
H₀: The median of the differences (D = X₂ – X₁) equals zero
H₁: The median of the differences ≠ 0 (two-tailed)
or > 0 or < 0 (one-tailed)
Step-by-Step Worked Example
Patient: 1 2 3 4 5 6 7 8
Before: 28 35 22 31 19 27 33 24
After: 21 30 25 22 18 19 28 20
Difference: -7 -5 +3 -9 -1 -8 -5 -4
1
State Hypotheses and Calculate Differences
H₀: Median anxiety difference (after − before) = 0. H₁: Median difference ≠ 0. Pairs with zero difference are dropped; effective n = count of non-zero differences.
2
Rank the Absolute Differences
Sorted |D|: 1, 3, 4, 5, 5, 7, 8, 9
Ranks: 1 2 3 4.5 4.5 6 7 8 (two 5s tied → average rank 4.5)
3
Assign Signs to Ranks and Calculate W
W+ (positive ranks) = 2
W- (negative ranks) = 6 + 4.5 + 8 + 1 + 7 + 4.5 + 3 = 34
W = min(W+, W-) = 2
Verification: W+ + W- = 36 = n(n+1)/2 = 8×9/2 = 36 ✓
4
Determine p-value and Effect Size
For n = 8 at α = 0.05 two-tailed, critical value = 4. Since W = 2 ≤ 4, reject H₀ (p < 0.05). The intervention significantly reduced anxiety.
z = (W – μ_W)/σ_W = (2 – 18)/7.14 ≈ -2.24
Effect size: r = |z|/√N = 2.24/√8 ≈ 0.79 (large)
Report: “A Wilcoxon Signed-Rank test indicated that anxiety scores were significantly lower following the intervention (Mdn = 21.5) compared to baseline (Mdn = 27.5), W = 2, z = -2.24, p = .025, r = .79.”
Why “Signed-Rank”? The name captures exactly what the test does. It uses both the sign (direction: positive or negative) and the rank (magnitude: how large the difference was relative to others) of each paired difference. This is what makes it more powerful than the simple sign test, which only uses direction.
Side by Side
Mann-Whitney U vs. Wilcoxon Signed-Rank: The Full Comparison
The two tests are often confused because both use ranks. The distinction is absolute: independent groups → Mann-Whitney U; paired groups → Wilcoxon Signed-Rank.
| Feature | Mann-Whitney U Test | Wilcoxon Signed-Rank Test |
|---|---|---|
| Also known as | Wilcoxon Rank-Sum Test, Mann-Whitney-Wilcoxon | Wilcoxon T Test, Paired Wilcoxon |
| Study design | Two independent groups (different subjects) | Two related/paired groups (same subjects or matched pairs) |
| Parametric equivalent | Independent samples t-test | Paired samples t-test |
| Null hypothesis | Both populations have the same distribution | Median of paired differences = 0 |
| Test statistic | U = min(U₁, U₂) | W = min(W+, W−) |
| What is ranked | All observations from both groups combined | Absolute values of differences between pairs |
| Effect size | r = |Z|/√N; rank-biserial correlation | r = |Z|/√N (same formula) |
| Software (SPSS) | Analyze → Nonparametric → 2 Independent Samples | Analyze → Nonparametric → 2 Related Samples |
| Software (R) | wilcox.test(x, y, paired = FALSE) | wilcox.test(x, y, paired = TRUE) |
Need Help Running or Interpreting These Tests?
Our statistics experts help with SPSS, R, Python, and hand calculations for Mann-Whitney U and Wilcoxon tests. Fast turnaround, step-by-step solutions.
Start an Order Login to AccountResults Reporting
Effect Sizes and How to Report Non-parametric Test Results
A p-value tells you whether your result is statistically significant. Effect size quantifies the magnitude of the difference, independent of sample size. Reporting effect sizes is required by APA Publication Manual (7th edition) and expected by journals and dissertation committees worldwide.
Effect size r = Z / √N
Where:
Z = the z-score from the normal approximation of the test statistic
N = total number of observations used in the test
Interpretation (Cohen’s benchmarks):
r = 0.10 → Small effect
r = 0.30 → Medium effect
r = 0.50 → Large effect
For the Mann-Whitney U, an alternative is the rank-biserial correlation:
r_rb = 1 – (2U) / (n₁ × n₂)
Ranges from -1 to +1; same benchmarks as r above.
Complete APA-Style Results Write-Up
Mann-Whitney U Test:
“A Mann-Whitney U test revealed a statistically significant difference in [DV]
between [Group 1] (Mdn = X) and [Group 2] (Mdn = X),
U = X, z = X, p = .XXX, r = .XX.”
Wilcoxon Signed-Rank Test:
“A Wilcoxon Signed-Rank test indicated that [DV] was significantly
[higher/lower] at [Condition 2] (Mdn = X) compared to [Condition 1] (Mdn = X),
W = X, z = X, p = .XXX, r = .XX.”
Reporting mistake to avoid: Do not report “the means were compared using a Mann-Whitney U test.” The Mann-Whitney U test does not compare means — it compares distributions. Always report medians, not means, as your central tendency measure for non-parametric tests.
Implementation
Running Non-parametric Tests in SPSS, R, and Python
SPSS
Mann-Whitney U in SPSS:
Analyze → Nonparametric Tests → Legacy Dialogs → 2 Independent Samples
1. Move DV to “Test Variable List”
2. Move grouping variable to “Grouping Variable” → Define Groups
3. Ensure “Mann-Whitney U” is checked
4. For small samples: click “Exact” → select “Exact”
5. Click OK → report U, Z, Asymp. Sig., and r = |Z|/√N
Wilcoxon Signed-Rank in SPSS:
Analyze → Nonparametric Tests → Legacy Dialogs → 2 Related Samples
1. Move both variables into the Test Pairs box
2. Ensure “Wilcoxon” is checked → Click OK
R
# Mann-Whitney U Test
wilcox.test(groupA, groupB, paired = FALSE, alternative = “two.sided”)
# Wilcoxon Signed-Rank Test
wilcox.test(after, before, paired = TRUE, alternative = “two.sided”)
# Note: R labels the paired statistic “V” — report it as W
Python
from scipy import stats
import numpy as np
# Mann-Whitney U Test
U_stat, p_value = stats.mannwhitneyu(group_a, group_b, alternative=’two-sided’)
# Wilcoxon Signed-Rank Test
W_stat, p_value = stats.wilcoxon(after, before, alternative=’two-sided’)
Exam Strategy
Common Mistakes and How to Avoid Them
Mistake 1: Confusing Independent and Paired Designs
Using Mann-Whitney U for paired data or Wilcoxon Signed-Rank for independent groups produces entirely invalid results. Always identify your design first: same subjects or matched pairs → Wilcoxon Signed-Rank; different people → Mann-Whitney U.
Mistake 2: Reporting Means Instead of Medians
Non-parametric tests operate on ranks. The appropriate measure of central tendency is the median, not the mean. Report Mdn = [value] for every non-parametric result.
Mistake 3: Omitting Effect Sizes
Reporting “p = 0.03” without effect size r is incomplete. Calculate r = |Z|/√N and include it in every reported result. APA 7th edition requires effect size reporting.
Mistake 4: Treating Non-parametric Tests as Assumption-Free
Independence of observations is critical — non-parametric tests fail just as badly as parametric ones when observations are correlated. Verify all three assumptions before presenting results.
Handling Ties
Tie correction for Mann-Whitney U z-statistic:
σ_U (corrected) = √[ n₁·n₂/12 × (N + 1 – Σtⱼ³/(N(N-1))) ]
where tⱼ = number of observations in the jth tied group
Software packages apply this correction automatically.
Quick test selection guide: Two groups, different people → Mann-Whitney U. Same people measured twice → Wilcoxon Signed-Rank. Three+ independent groups → Kruskal-Wallis. Three+ related groups → Friedman test. Non-parametric correlation → Spearman’s ρ.
Statistics Assignment or Dissertation Analysis Due?
From test selection and SPSS/R output interpretation to full write-ups with effect sizes — our statistics experts deliver fast, reliable, explained solutions for students at every level.
Get Help With My Assignment Log InFrequently Asked
Frequently Asked Questions About Non-parametric Tests
What is a non-parametric test and when should I use one?
A non-parametric test is a statistical hypothesis test that does not require data to follow a specific distribution. Non-parametric tests rank observations rather than using raw values, making them robust to non-normality, outliers, and ordinal measurement. Use one when your data is ordinal (Likert scales, rankings), when your continuous data is significantly non-normal with small samples (n < 30), when you have extreme outliers, or when your data contains many ties.
What is the Mann-Whitney U test used for?
The Mann-Whitney U test compares two independent groups to determine whether one group tends to have systematically higher or lower values than the other. It is the non-parametric alternative to the independent samples t-test. Use it when your two groups consist of different, unrelated subjects and your dependent variable is ordinal or non-normally distributed.
What is the difference between Mann-Whitney U and Wilcoxon Signed-Rank?
The fundamental difference is your research design. Mann-Whitney U is for two independent groups — different people in each group. Wilcoxon Signed-Rank is for two related/paired groups — the same people measured under both conditions, or carefully matched pairs. Both tests use ranks, but applying the wrong test invalidates your results entirely.
How do I calculate the effect size for non-parametric tests?
Calculate effect size r = |Z| / √N, where Z is the z-score from the test and N is the total number of observations. Benchmarks: r = 0.1 (small), r = 0.3 (medium), r = 0.5 (large). Always report effect sizes alongside p-values — statistical significance tells you an effect exists, effect size tells you whether it matters practically.
Do non-parametric tests have any assumptions?
Yes — non-parametric tests are not assumption-free. The Mann-Whitney U requires: (1) independent observations, (2) at least ordinal measurement, (3) similarly shaped distributions if comparing medians. The Wilcoxon Signed-Rank requires: (1) paired observations, (2) continuous or ordinal measurement, (3) symmetrically distributed differences. Normality is NOT required by either test.
Can I use a Mann-Whitney U test for Likert scale data?
Yes — the Mann-Whitney U test is appropriate for Likert scale data because it only requires ordinal measurement. Likert data has ordered categories but not guaranteed equal intervals, which violates the continuous measurement assumption of the t-test. For most university-level work, non-parametric tests for individual Likert items is the safer and more defensible choice.
How do I report non-parametric test results in APA format?
APA 7th edition format: “A Mann-Whitney U test revealed a significant difference between [Group 1] (Mdn = X) and [Group 2] (Mdn = X), U = X, z = X, p = .XXX, r = .XX.” Key rules: (1) always report medians, not means; (2) always include effect size r; (3) report exact p-values; (4) use lowercase italics for statistical symbols.
Is the Mann-Whitney U test the same as the Wilcoxon Rank-Sum test?
Yes — the Mann-Whitney U test and the Wilcoxon Rank-Sum test are mathematically equivalent procedures for comparing two independent groups, producing identical p-values. The difference is only in how the test statistic is expressed. Do not confuse this with the Wilcoxon Signed-Rank test, which is for paired data — they are different tests despite sharing “Wilcoxon” in the name.