Study guide
9.3 Paired t-Tests & Non-Parametric Alternatives
CSSC Certified Six Sigma Green Belt Exam
Track your progress as you read
Finished sections get a check mark.
1 Chapter 1: Six Sigma Overview & Exam Fundamentals
1.1 CSSC Certification Overview, Exam Structure & Body of Knowledge 1.2 History & Evolution of Six Sigma 1.3 Six Sigma Organizational Roles & Belts Hierarchy 1.4 The DMAIC Methodology Architecture vs. DMADV / DFSS 1.5 Other Process Improvement & Quality Methods, and When to Choose Six Sigma
2 Chapter 2: Quality Fundamentals & Defect Metrics
2.1 Critical to Quality (CTQ) Characteristics & VOC Translation 2.2 Defect Metrics: DPU, DPO, DPMO, & the 1.5-Sigma Shift 2.3 First Time Yield (FTY) & Rolled Throughput Yield (RTY) 2.4 Cost of Quality (COQ) & Cost of Poor Quality (COPQ)
3 Chapter 3: Lean Principles & Waste Elimination
3.1 Foundations of Lean & The Toyota Production System (TPS) 3.2 The Eight Wastes: CSSC's Seven Muda, TIMWOODS, and Muri/Mura 3.3 The 5S Visual Workplace Methodology 3.4 Value Stream Mapping (VSM), Cycle Time & Takt Time
4 Chapter 4: Project Selection & Team Management
4.1 Six Sigma Project Selection & Business Case Development 4.2 Project Charter Architecture 4.3 Stakeholder Management & Communication Planning 4.4 Team Dynamics & Tuckman's Team Development Stages 4.5 The 15-Point Project Viability Model & Process-Level Selection
5 Chapter 5: Define Phase & Process Mapping
5.1 High-Level Process Mapping with SIPOC 5.2 Flowcharts, Swimlane Diagrams & Process Value Analysis 5.3 Quality Function Deployment (QFD) & The House of Quality 5.4 Critical to Quality (CTQ) Flowdown & Operational Definitions 5.5 The CSSC Define Toolset: Ground Rules, In/Out of the Box, Is–Is Not & Tollgate Checklists
6 Chapter 6: Measure Phase & Measurement Systems
6.1 Data Types: Continuous (Variable) vs. Discrete (Attribute) Data 6.2 Data Collection Planning & Sampling Strategies 6.3 Measurement System Analysis (MSA) Foundations & Calibration 6.4 Variable & Attribute Gage R&R Studies
7 Chapter 7: Analyze Phase & Root Cause Analysis
7.1 Qualitative Root Cause Tools: 5 Whys & Cause-and-Effect (Fishbone) Diagrams 7.2 Failure Mode and Effects Analysis (FMEA) & RPN Calculation 7.3 Pareto Analysis & The 80/20 Rule 7.4 Graphical Data Exploration: Histograms, Box Plots, & Multi-Vari Charts 7.5 The Root Cause Verification Matrix
8 Chapter 8: Descriptive Statistics & Distributions
8.1 Measures of Central Tendency & Dispersion 8.2 The Normal Distribution & The Empirical Rule (68-95-99.7) 8.3 Standard Normal Distribution, Z-Scores & Probability Calculations 8.4 Common Non-Normal Distributions: Binomial, Poisson & Chi-Square
9 Chapter 9: Inferential Statistics & Hypothesis Testing
9.1 Fundamentals of Hypothesis Testing: Null vs. Alternative, Alpha & Beta Risks 9.2 One-Sample and Two-Sample t-Tests for Means 9.3 Paired t-Tests & Non-Parametric Alternatives 9.4 Analysis of Variance (One-Way & Two-Way ANOVA) & Chi-Square Contingency Tests 9.5 Sample Size Determination for Hypothesis Tests
10 Chapter 10: Correlation, Regression, & Experimentation
10.1 Linear Correlation & Pearson's Correlation Coefficient (r) 10.2 Simple Linear Regression & Coefficient of Determination (R^2) 10.3 Introduction to Design of Experiments (DOE) & Factorial Designs 10.4 Analyzing Experimental Results: Main Effects & Interaction Plots
11 Chapter 11: Improve Phase & Solution Implementation
11.1 Brainstorming, Creative Thinking & Solution Selection Matrices 11.2 Piloting Solutions & Conducting Risk Assessments 11.3 Implementation Planning & Change Management (ADKAR) 11.4 Mistake-Proofing (Poka-Yoke) & Standard Work
12 Chapter 12: Control Phase, SPC, & Process Capability
12.1 Statistical Process Control (SPC) & Shewhart Control Rules 12.2 Variable Control Charts: I-MR, X-bar & R, X-bar & S 12.3 Attribute Control Charts: p, np, c, and u Charts 12.4 Process Capability & Performance Indices: Cp, Cpk, Pp, Ppk 12.5 Communicating Statistical Results to Business Audiences
9.3 Paired t-Tests & Non-Parametric Alternatives
Key Takeaways Explain This Topic AI
The Paired t-test evaluates dependent, matched, or repeated measurements taken on identical experimental units (e.g., Before vs. After), eliminating extraneous unit-to-unit background variation.
Mathematically, the Paired t-test is identical to a One-Sample t-test performed on difference scores (di = x1i - x2i), with degrees of freedom df = n_pairs - 1.
By isolating within-unit changes from between-unit differences, pairing drastically shrinks the standard error and maximizes statistical power compared to an independent two-sample design.
When continuous data severely violates normality, resists mathematical transformation (e.g., Box-Cox), or consists of ordinal ranks, non-parametric tests evaluate medians and ranked data.
Parametric tests map directly to non-parametric equivalents: Two-Sample t-test to Mann-Whitney U, Paired t-test to Wilcoxon Signed-Rank, One-Way ANOVA to Kruskal-Wallis or Mood's Median, and Pearson's r to Spearman's ρ.
Last updated: September 2026
9.3 Paired t-Tests & Non-Parametric Alternatives
Quick Summary: While independent two-sample tests compare separate groups, many continuous improvement studies involve dependent, repeated, or matched observations—such as evaluating the exact same machines Before and After a maintenance kaizen, or having two appraisers inspect identical components. The Paired t t t -test isolates within-unit changes by computing difference scores ( d i = x 1 i − x 2 i d_i = x_{1i} - x_{2i} d i = x 1 i − x 2 i ), subtracting out background part-to-part variation and dramatically boosting statistical power . When continuous data severely violates the Gaussian normality assumption and cannot be transformed, practitioners deploy non-parametric tests (Mann-Whitney U, Wilcoxon Signed-Rank, Kruskal-Wallis) to test medians and ranked data rather than means.
The Engineering Power of Dependent (Paired) Sampling
In standard manufacturing, healthcare, and administrative processes, natural unit-to-unit variation is often massive:
In a machining study, casting porosity and alloy density vary naturally from block to block.
In a clinical study, metabolic rates and blood chemistry vary naturally from patient to patient.
In loan underwriting, applicant financial history varies widely across applications.
If an engineering team tests a process improvement using an Independent Two-Sample t t t -test —measuring 20 parts made with Setup A and 20 completely different parts made with Setup B—the huge natural casting-to-casting variation enters the error term ( s p s_p s p ). The signal of the improvement is drowned out by the background noise of the parts, causing an underpowered test and a Type II Error (missed detection).
How Pairing Eliminates Background Noise
By taking paired measurements on the exact same physical units under both conditions (or using identically matched twins), the background unit-to-unit variation is mathematically subtracted out: How Pairing Eliminates Background Noise
- INDEPENDENT SAMPLES PAIRED SAMPLES
- ┌─────────────────────────────┐ ┌─────────────────────────────┐
- │ Shift A parts vs. │ │ Same parts measured BEFORE │
- │ Shift B parts │ │ and AFTER maintenance │
- ├─────────────────────────────┤ ├─────────────────────────────┤
- │ High part-to-part variation │ │ Part variation cancels out: │
- │ inflates standard error! │ │ di = x_After - x_Before │
- │ Lower Statistical Power │ │ High Statistical Power │
- └─────────────────────────────┘ └─────────────────────────────┘
Common Six Sigma paired applications include:
Before vs. After Studies: Measuring cycle times, temperature, or emissions on the identical set of machines before and after a Lean kaizen event.
Measurement System Analysis (MSA): Appraiser 1 and Appraiser 2 measuring the identical 20 production samples to evaluate appraiser bias.
Split-Lot / Matched Testing: Applying Coating Formulation A to one half of a silicon wafer and Coating Formulation B to the other half of the identical wafer.
Mathematical Formulation of the Paired t-Test
The Paired t t t -test is mathematically identical to a One-Sample t t t -test performed on the single column of difference scores ( d i d_i d i ): d i = x 1 i − x 2 i d_i = x_{1i} - x_{2i} d i = x 1 i − x 2 i
Hypotheses Formulation
Null Hypothesis ( H 0 H_0 H 0 ): μ d = Δ 0 \mu_d = \Delta_0 μ d = Δ 0 (Typically H 0 : μ d = 0 H_0: \mu_d = 0 H 0 : μ d = 0 , asserting zero mean difference between paired conditions).
Alternative Hypothesis ( H a H_a H a ): μ d ≠ 0 \mu_d \ne 0 μ d = 0 (Two-tailed), μ d > 0 \mu_d > 0 μ d > 0 , or μ d < 0 \mu_d < 0 μ d < 0 (One-tailed).
Test Statistic & Formulas t calc = d ˉ − Δ 0 s d / n t_{\text{calc}} = \frac{\bar{d} - \Delta_0}{s_d / \sqrt{n}} t calc = s d / n d ˉ − Δ 0
Where:
d ˉ = ∑ i = 1 n d i n = \bar{d} = \frac{\sum_{i=1}^n d_i}{n} = d ˉ = n ∑ i = 1 n d i = Arithmetic mean of the differences
s d = ∑ i = 1 n ( d i − d ˉ ) 2 n − 1 = s_d = \sqrt{\frac{\sum_{i=1}^n (d_i - \bar{d})^2}{n - 1}} = s d = n − 1 ∑ i = 1 n ( d i − d ˉ ) 2 = Sample standard deviation of the differences
- n = n = n = Number of pairs (NOT the total number of raw observations 2 n 2n 2 n )
- S E d = s d n = SE_d = \frac{s_d}{\sqrt{n}} = S E d = n s d = Standard error of the mean difference
D e g r e e s o f F r e e d o m : d f = n − 1 \mathbf{Degrees\ of\ Freedom:}\ df = n - 1 Degrees of Freedom : df = n − 1
Exam Watchout: If 15 parts are measured Before and After (giving 30 total measurements), the degrees of freedom is d f = 15 − 1 = 14 df = 15 - 1 = 14 df = 15 − 1 = 14 . A common distractor on the CSSC exam is d f = 30 − 2 = 28 df = 30 - 2 = 28 df = 30 − 2 = 28 , which erroneously applies the independent two-sample formula.
Step-by-Step Worked Example: Paired t-Test
Scenario: A Six Sigma Green Belt at an electronics manufacturing plant designs a new ergonomic hand fixture to reduce wiring harness assembly cycle time. The team times n = 10 n = 10 n = 10 assembly technicians using the baseline standard fixture ( T Base T_{\text{Base}} T Base ) and the new ergonomic fixture ( T New T_{\text{New}} T New ). Assembly times are recorded in seconds:
Operator ( i i i ) Baseline Time T Base T_{\text{Base}} T Base (s) Ergonomic Time T New T_{\text{New}} T New (s) Difference d i = T New − T Base d_i = T_{\text{New}} - T_{\text{Base}} d i = T New − T Base ( d i − d ˉ ) (d_i - \bar{d}) ( d i − d ˉ ) ( d i − d ˉ ) 2 (d_i - \bar{d})^2 ( d i − d ˉ ) 2
- 1 54.2 49.8 -4.4 -0.20 0.040
- 2 62.8 57.5 -5.3 -1.10 1.210
- 3 48.5 45.1 -3.4 +0.80 0.640
- 4 71.0 66.2 -4.8 -0.60 0.360
- 5 55.4 52.0 -3.4 +0.80 0.640
- 6 68.2 62.0 -6.2 -2.00 4.000
- 7 50.1 47.3 -2.8 +1.40 1.960
- 8 58.7 54.9 -3.8 +0.40 0.160
- 9 65.3 61.1 -4.2 0.00 0.000
- 10 59.8 56.1 -3.7 +0.50 0.250
- Sum 594.0 552.0 -42.0 0.00 ∑ = 9.260 \sum = 9.260 ∑ = 9.260
Test whether the new ergonomic fixture achieves a statistically significant reduction in assembly cycle time at α = 0.05 \alpha = 0.05 α = 0.05 .
- Step 1: State the Hypotheses
- H 0 : μ d ≥ 0 H_0: \mu_d \ge 0 H 0 : μ d ≥ 0 (New fixture does not reduce assembly cycle time)
H a : μ d < 0 H_a: \mu_d < 0 H a : μ d < 0 (New fixture significantly reduces assembly cycle time; one-tailed lower test)
Step 2: Compute Mean and Standard Deviation of Differences
d ˉ = ∑ d i n = − 42.0 10 = − 4.20 seconds \bar{d} = \frac{\sum d_i}{n} = \frac{-42.0}{10} = -4.20\text{ seconds} d ˉ = n ∑ d i = 10 − 42.0 = − 4.20 seconds
s d = ∑ ( d i − d ˉ ) 2 n − 1 = 9.260 10 − 1 = 9.260 9 = 1.0289 = 1.0143 seconds s_d = \sqrt{\frac{\sum (d_i - \bar{d})^2}{n - 1}} = \sqrt{\frac{9.260}{10 - 1}} = \sqrt{\frac{9.260}{9}} = \sqrt{1.0289} = 1.0143\text{ seconds} s d = n − 1 ∑ ( d i − d ˉ ) 2 = 10 − 1 9.260 = 9 9.260 = 1.0289 = 1.0143 seconds
Step 3: Calculate Standard Error and Test Statistic ( t calc t_{\text{calc}} t calc )
Standard Error: S E d = s d n = 1.0143 10 = 1.0143 3.1623 = 0.3208 seconds SE_d = \frac{s_d}{\sqrt{n}} = \frac{1.0143}{\sqrt{10}} = \frac{1.0143}{3.1623} = 0.3208\text{ seconds} S E d = n s d = 10 1.0143 = 3.1623 1.0143 = 0.3208 seconds
t calc = d ˉ − 0 S E d = − 4.20 − 0 0.3208 = − 13.09 t_{\text{calc}} = \frac{\bar{d} - 0}{SE_d} = \frac{-4.20 - 0}{0.3208} = -13.09 t calc = S E d d ˉ − 0 = 0.3208 − 4.20 − 0 = − 13.09
- Step 4: Make the Decision
- Degrees of freedom: d f = n − 1 = 10 − 1 = 9 df = n - 1 = 10 - 1 = 9 df = n − 1 = 10 − 1 = 9
Critical value from t t t -table for one-tailed lower test at α = 0.05 \alpha = 0.05 α = 0.05 with d f = 9 df = 9 df = 9 : t crit = − 1.833 t_{\text{crit}} = -1.833 t crit = − 1.833
Since t calc = − 13.09 < − 1.833 t_{\text{calc}} = -13.09 < -1.833 t calc = − 13.09 < − 1.833 (software p < 0.0001 p < 0.0001 p < 0.0001 ), we Reject H 0 H_0 H 0 .
Conclusion: The new ergonomic fixture achieves an extraordinary, statistically significant cycle time reduction of 4.20 seconds 4.20\text{ seconds} 4.20 seconds per harness ( p < 0.0001 p < 0.0001 p < 0.0001 ).
When Parametric Assumptions Fail: The Non-Parametric Domain
Parametric tests ( t t t -tests, ANOVA) rely strictly on the assumption that underlying continuous distributions are approximately Gaussian normal. In Six Sigma, three common conditions violate this assumption:
Severe Skewness / Asymmetry: Customer wait times, repair cycle times, and pollutant concentrations often exhibit extreme right-skewed lognormal or exponential shapes.
Small Sample Sizes with Outliers ( n < 30 n < 30 n < 30 ): When sample sizes are small, the Central Limit Theorem cannot be invoked to ensure normality of sample means, and extreme outliers severely distort x ˉ \bar{x} x ˉ and s s s .
Ordinal / Ranked Data: Customer satisfaction ratings (1 to 5 stars), pain scales, or supplier audit tier ranks are ordinal attributes, not true continuous interval metrics.
The Remediation Hierarchy
When data violates normality, Green Belts follow a structured three-step escalation protocol:
Check for Data Errors: Investigate whether extreme outliers are typos or measurement errors.
Attempt Mathematical Transformation: Apply Box-Cox or logarithmic power transformations to convert non-normal data into approximate normality.
Deploy Non-Parametric Tests: If data resists transformation or sample sizes are small, execute non-parametric tests .
Non-Parametric Tests for Medians & Ranks
Non-parametric tests—frequently termed distribution-free tests —make zero assumptions regarding the specific mathematical shape of the population. Instead of evaluating means and variances, non-parametric tests evaluate medians ( μ ~ \tilde{\mu} μ ~ or η \eta η ) and analyze the ranks of ordered observations. Parametric vs. Non-Parametric Test Mapping
- PARAMETRIC TEST (Means) NON-PARAMETRIC EQUIVALENT (Medians)
- ┌─────────────────────────────────────┐ ┌─────────────────────────────────────┐
- │ One-Sample t-Test │ ──▶│ 1-Sample Wilcoxon / 1-Sample Sign │
- ├─────────────────────────────────────┤ ├─────────────────────────────────────┤
- │ Two-Sample Independent t-Test │ ──▶│ Mann-Whitney U / Wilcoxon Rank-Sum │
- ├─────────────────────────────────────┤ ├─────────────────────────────────────┤
- │ Paired t-Test │ ──▶│ Wilcoxon Signed-Rank Test │
- ├─────────────────────────────────────┤ ├─────────────────────────────────────┤
- │ One-Way ANOVA │ ──▶│ Kruskal-Wallis / Mood's Median │
- ├─────────────────────────────────────┤ ├─────────────────────────────────────┤
- │ Pearson Correlation (r) │ ──▶│ Spearman's Rank Correlation (ρ) │
- └─────────────────────────────────────┘ └─────────────────────────────────────┘
- 1. The Mann-Whitney U Test (Wilcoxon Rank-Sum Test)
Parametric Equivalent: Two-Sample Independent t t t -test.
Objective: Compares the medians of two independent, unrelated groups ( n 1 n_1 n 1 and n 2 n_2 n 2 ).
Mechanics: All observations from both groups are pooled into a single combined dataset and ranked in ascending order from 1 1 1 to N = n 1 + n 2 N = n_1 + n_2 N = n 1 + n 2 . Tied values receive average ranks. The sum of ranks for each group is calculated ( W 1 W_1 W 1 and W 2 W_2 W 2 ). If the two populations are identical, the average rank in both groups should be equal.
Test Statistic ( U U U ): U 1 = n 1 n 2 + n 1 ( n 1 + 1 ) 2 − W 1 U_1 = n_1 n_2 + \frac{n_1(n_1 + 1)}{2} - W_1 U 1 = n 1 n 2 + 2 n 1 ( n 1 + 1 ) − W 1 U 2 = n 1 n 2 + n 2 ( n 2 + 1 ) 2 − W 2 U_2 = n_1 n_2 + \frac{n_2(n_2 + 1)}{2} - W_2 U 2 = n 1 n 2 + 2 n 2 ( n 2 + 1 ) − W 2
Hypotheses: H 0 : μ ~ 1 = μ ~ 2 H_0: \tilde{\mu}_1 = \tilde{\mu}_2 H 0 : μ ~ 1 = μ ~ 2 vs H a : μ ~ 1 ≠ μ ~ 2 H_a: \tilde{\mu}_1 \ne \tilde{\mu}_2 H a : μ ~ 1 = μ ~ 2 .
2. The Wilcoxon Signed-Rank Test (Paired)
Parametric Equivalent: Paired t t t -test.
Objective: Compares dependent, matched pairs when difference scores are non-normal but roughly symmetric.
Mechanics: Differences d i = x 1 i − x 2 i d_i = x_{1i} - x_{2i} d i = x 1 i − x 2 i are calculated. Any pairs where d i = 0 d_i = 0 d i = 0 are discarded. The absolute values ∣ d i ∣ |d_i| ∣ d i ∣ are ranked from smallest to largest. The original positive or negative sign is reattached to each rank. The test sums the positive ranks ( W + W^+ W + ) and negative ranks ( W − W^- W − ) and evaluates whether their sum deviates significantly from expected random variation.
3. The 1-Sample Sign Test
Parametric Equivalent: One-Sample t t t -test.
Objective: Tests whether a population median equals a specified target ( μ ~ = M 0 \tilde{\mu} = M_0 μ ~ = M 0 ).
Mechanics: Extremely simple: counts how many observations fall strictly above ( + + + ) and below ( − - − ) the hypothesized target M 0 M_0 M 0 . Under H 0 H_0 H 0 , the probability of an observation falling above the median is p = 0.50 p = 0.50 p = 0.50 . The test uses the Binomial distribution to compute exact p p p -values. Highly robust against extreme outliers.
4. Kruskal-Wallis Test & Mood's Median Test
Parametric Equivalent: One-Way ANOVA.
Kruskal-Wallis: Compares medians across k ≥ 3 k \ge 3 k ≥ 3 independent groups by ranking all combined observations across groups. Sensitive to both shifts in median and general distribution shape.
Mood's Median Test: Tests whether k ≥ 3 k \ge 3 k ≥ 3 groups share a common median by determining the grand median across all data, constructing a 2 × k 2 \times k 2 × k contingency table (counts above vs. below grand median), and running a Chi-Square test. Mood's median test is vastly more resistant to extreme outliers than Kruskal-Wallis, though it possesses slightly lower statistical power.
Parametric vs. Non-Parametric Selection Tradeoffs
Dimension Parametric Tests ( t t t -Tests, ANOVA) Non-Parametric Tests (Mann-Whitney, Kruskal-Wallis)
Metric Evaluated Means ( μ \mu μ ) and standard deviations ( σ \sigma σ ) Medians ( μ ~ \tilde{\mu} μ ~ ) and rank positions
Required Distribution Gaussian Normal (or n ≥ 30 n \ge 30 n ≥ 30 via CLT) Distribution-free (no shape assumption)
Data Measurement Scale Continuous Interval or Ratio Continuous or Ordinal Ranks (Likert scales)
Outlier Sensitivity High (outliers inflate s s s and distort x ˉ \bar{x} x ˉ ) Low (resistant; ranks neutralize extreme values)
Statistical Power Maximum power when normality holds Lower power (Asymptotic Relative Efficiency ~95% of t t t -test)
Sample Size Needs Efficient with small samples if normal Requires slightly larger sample sizes to match power
The Six Sigma Rule of Thumb: If data is normal, always use parametric tests because they maximize statistical power. If data is severely non-normal, resists transformation, or consists of ordinal categories, deploy non-parametric tests to prevent invalid conclusions.
Critical CSSC Exam Traps
Trap 1: Running an Independent Two-Sample Test on Paired Data — If you run an independent Two-Sample t t t -test on Before/After measurements, the test treats the 2 n 2n 2 n data points as unrelated, failing to subtract out unit-to-unit variation. The standard error balloons, causing a Type II Error (failing to detect a real improvement).
Trap 2: Degrees of Freedom in Paired Studies — For a paired test with 12 pairs measured twice (24 total observations), d f = n pairs − 1 = 12 − 1 = 11 df = n_{\text{pairs}} - 1 = 12 - 1 = 11 df = n pairs − 1 = 12 − 1 = 11 , NOT 24 − 2 = 22 24 - 2 = 22 24 − 2 = 22 .
Trap 3: Believing Non-Parametric Tests Evaluate Means — Non-parametric tests evaluate medians or rank sums, never arithmetic means.
Trap 4: Mood's Median vs. Kruskal-Wallis Selection — Kruskal-Wallis is preferred for general non-normal data across ≥ 3 \ge 3 ≥ 3 groups. However, if extreme outliers or unequal subgroup shapes are present, Mood's Median test is the superior choice because it is completely unaffected by outlier magnitudes.