Study guide
7.5 Advanced Statistical Design & Analysis
Certified Quality Engineer CQE
Track your progress as you read
Finished sections get a check mark.
1 Chapter 1: Management and Leadership
1.1 Quality Philosophies & Foundations 1.2 Leadership & Organizational Systems 1.3 ASQ Code of Ethics & Professional Standards 1.4 Strategic Planning & Deployment 1.5 Management Infrastructure & Project Management 1.6 Quality Information Systems & Communication 1.7 Financial & Cost of Quality
2 Chapter 2: The Quality System
2.1 Quality Management System Elements & Standards 2.2 Quality Documentation Architecture 2.3 Quality Auditing Principles & Types 2.4 Audit Planning, Execution & Reporting 2.5 Corrective Action & Preventive Action Systems 2.6 Supplier Quality Management 2.7 Customer Relations & Customer Value 2.8 Quality Training Program Design & Effectiveness
3 Chapter 3: Product, Process, and Service Design
3.1 Design Input & Requirements 3.2 Design Review & Verification 3.3 Failure Mode and Effects Analysis 3.4 Design of Experiments Foundations 3.5 Robust Design & Taguchi Methods 3.6 Reliability & Maintainability Concepts 3.7 Reliability Modeling & System Evaluation 3.8 Technical Drawings, Specifications & GD&T
4 Chapter 4: Product and Process Control
4.1 Statistical Process Control Principles 4.2 Variable Control Charts 4.3 Attribute Control Charts 4.4 Process Capability & Performance Analysis 4.5 Acceptance Sampling Plans & Standards 4.6 Metrology, Calibration & Measurement Uncertainty 4.7 Measurement System Analysis (MSA) 4.8 Material Control, Traceability & MRB
5 Chapter 5: Continuous Improvement
5.1 Continuous Improvement Methodologies 5.2 Classic Quality Tools 5.3 Management & Planning Tools 5.4 Lean Tools & Waste Elimination 5.5 Problem-Solving Methodologies & Root Cause Analysis 5.6 Risk Management & Assessment 5.7 Change Management & Team Facilitation
6 Chapter 6: Quantitative Methods and Tools
6.1 Descriptive Statistics & Data Summarization 6.2 Probability Concepts & Distributions 6.3 Sampling Distributions & Central Limit Theorem 6.4 Point & Interval Estimation 6.5 Hypothesis Testing Fundamentals 6.6 Specific Statistical Tests 6.7 Regression & Correlation Analysis
7 Chapter 7: Risk Management & Advanced Quality Engineering
7.1 Risk Identification & Hazard Analysis 7.2 Quantitative Risk Analysis & Modeling 7.3 Risk Mitigation & Control Planning 7.4 Special Process Control & Advanced SPC 7.5 Advanced Statistical Design & Analysis 7.6 Computer Applications & Quality Engineering Software
7.5 Advanced Statistical Design & Analysis
Key Takeaways Explain This Topic AI
Fractional Factorial 2^(k-p) designs reduce experimental runs by selecting generators (e.g., I = ABC), introducing confounding patterns described by the defining relation and aliasing structure.
Design Resolution defines the shortest word length in the defining relation: Resolution III aliases main effects with 2-factor interactions, Resolution IV aliases 2-factor interactions with each other, and Resolution V aliases 2-factor interactions with 3-factor interactions.
Response Surface Methodology (RSM) utilizes Central Composite Designs (CCD) with 2^k factorial points, 2k axial (star) points at distance α, and n_c center points to fit full quadratic response models.
Rotatability in CCD requires α = (N_F)^(1/4) (e.g., α = 1.414 for k=2, α = 1.682 for k=3), ensuring equal prediction variance at all points equidistant from the design center.
Last updated: July 2026
7.5 Advanced Statistical Design & Analysis
Design of Experiments (DOE) is a structured statistical methodology used to manipulate input process factors ( X X X ) systematically and evaluate their effect on key response variables ( Y Y Y ). While full factorial designs evaluate all possible factor combinations, resource constraints in complex manufacturing often demand advanced DOE techniques.
This section covers Fractional Factorial Designs ( 2 k − p 2^{k-p} 2 k − p ) , Analysis of Variance (ANOVA) for DOE , and Response Surface Methodology (RSM) using Central Composite and Box-Behnken designs.
1. Fractional Factorial Designs ( 2 k − p 2^{k-p} 2 k − p )
When the number of factors k k k increases, the number of treatment combinations in a full 2 k 2^k 2 k factorial design grows exponentially (e.g., 2 6 = 64 2^6 = 64 2 6 = 64 runs, 2 7 = 128 2^7 = 128 2 7 = 128 runs). According to the Sparsity of Effects Principle (or Pareto principle in DOE), a system is primarily driven by main effects and low-order (2-factor) interactions. High-order interactions (3-factor or higher) are rarely statistically significant.
A Fractional Factorial design , denoted as 2 k − p 2^{k-p} 2 k − p , executes a fraction ( 1 / 2 p 1/2^p 1/ 2 p ) of the full factorial runs by setting p p p factors equal to high-order interaction columns (called generators ).
Defining Relations and Aliasing Structure
When factor columns are set equal to interaction columns, information becomes confounded (or aliased ). Two effects are aliased if their calculated contrast estimates measure the exact same numerical quantity.
Example: 2 3 − 1 2^{3-1} 2 3 − 1 Half-Fraction Design
Factors: A , B , C A, B, C A , B , C ( k = 3 k=3 k = 3 ), Fraction: p = 1 p=1 p = 1 ( 1 / 2 1/2 1/2 fraction = 4 4 4 runs).
Generator choice: C = A B C = AB C = A B .
Multiplying both sides by C C C yields the Defining Relation : I = A B C I = ABC I = A BC
To determine the aliasing of any effect, multiply that effect by the defining relation (using modulo-2 arithmetic where A 2 = I A^2 = I A 2 = I ): A ⋅ I = A ⋅ A B C ⟹ A = B C A \cdot I = A \cdot ABC \implies A = BC A ⋅ I = A ⋅ A BC ⟹ A = BC B ⋅ I = B ⋅ A B C ⟹ B = A C B \cdot I = B \cdot ABC \implies B = AC B ⋅ I = B ⋅ A BC ⟹ B = A C C ⋅ I = C ⋅ A B C ⟹ C = A B C \cdot I = C \cdot ABC \implies C = AB C ⋅ I = C ⋅ A BC ⟹ C = A B
Thus, in a 2 3 − 1 2^{3-1} 2 3 − 1 design with I = A B C I=ABC I = A BC , main effect A A A is aliased with interaction B C BC BC , B B B with A C AC A C , and C C C with A B AB A B .
Design Resolution Classification
Design Resolution (denoted by Roman numerals III, IV, V) characterizes the degree of aliasing in a fractional factorial design. The resolution equals the length of the shortest word in the defining relation I I I .
Resolution Shortest Word Length Aliasing Rule Application / Capability
Resolution III 3 (e.g., I = A B C I = ABC I = A BC ) Main effects are aliased with 2-factor interactions. Screening Design : Fast identification of active main effects; assumes 2-factor interactions are negligible.
Resolution IV 4 (e.g., I = A B C D I = ABCD I = A BC D ) Main effects are unaliased with 2-factor interactions, but 2-factor interactions are aliased with each other. Characterization Design : Main effects are clear; requires foldover to resolve confounded 2-factor interactions.
Resolution V 5 (e.g., I = A B C D E I = ABCDE I = A BC D E ) Main effects and 2-factor interactions are unaliased with each other. 2-factor interactions are aliased with 3-factor interactions. Optimization Design : Excellent resolution of all main effects and 2-factor interactions without full factorial cost.
2. Analysis of Variance (ANOVA) Models for DOE
ANOVA partitions the total variation observed in the experimental response ( S S Total SS_{\text{Total}} S S Total ) into components attributable to specific design factors, interactions, and random experimental error ( S S Error SS_{\text{Error}} S S Error ).
Fundamental ANOVA Equation S S Total = S S Model + S S Error = ∑ i = 1 k S S Factor i + ∑ S S Interactions + S S Error SS_{\text{Total}} = SS_{\text{Model}} + SS_{\text{Error}} = \sum_{i=1}^k SS_{\text{Factor}_i} + \sum SS_{\text{Interactions}} + SS_{\text{Error}} S S Total = S S Model + S S Error = i = 1 ∑ k S S Factor i + ∑ S S Interactions + S S Error
Degrees of Freedom and Mean Squares
For an experiment with N N N total observations and p p p estimated parameters:
Total Degrees of Freedom: d f Total = N − 1 df_{\text{Total}} = N - 1 d f Total = N − 1
Factor Degrees of Freedom (for a 2-level factor): d f Factor = 2 − 1 = 1 df_{\text{Factor}} = 2 - 1 = 1 d f Factor = 2 − 1 = 1
Error Degrees of Freedom: d f Error = N − p df_{\text{Error}} = N - p d f Error = N − p
The Mean Square ( M S MS MS ) for any term is calculated by dividing its Sum of Squares ( S S SS SS ) by its degrees of freedom ( d f df df ): M S Effect = S S Effect d f Effect , M S Error = S S Error d f Error MS_{\text{Effect}} = \frac{SS_{\text{Effect}}}{df_{\text{Effect}}}, \quad MS_{\text{Error}} = \frac{SS_{\text{Error}}}{df_{\text{Error}}} M S Effect = d f Effect S S Effect , M S Error = d f Error S S Error
Test Statistic ( F 0 F_0 F 0 ) and Hypothesis Testing
To test the null hypothesis H 0 : β Effect = 0 H_0: \beta_{\text{Effect}} = 0 H 0 : β Effect = 0 , the F F F -ratio is computed: F 0 = M S Effect M S Error F_0 = \frac{MS_{\text{Effect}}}{MS_{\text{Error}}} F 0 = M S Error M S Effect
If F 0 > F α , d f Effect , d f Error F_0 > F_{\alpha, df_{\text{Effect}}, df_{\text{Error}}} F 0 > F α , d f Effect , d f Error (or p -value < α p\text{-value} < \alpha p -value < α ), the factor effect is statistically significant. Source of Variation | Sum of Squares (SS) | Degrees of Freedom (df) | Mean Square (MS) | F-Ratio (F0)
- --------------------------------------------------------------------------------------------------
- Factor A | SSA | 1 | MSA = SSA / 1 | MSA / MSE
- Factor B | SSB | 1 | MSB = SSB / 1 | MSB / MSE
- Interaction AB | SSAB | 1 | MSAB = SSAB / 1 | MSAB / MSE
- Residual Error | SSE | N - 4 | MSE = SSE / dfE | -
- Total | SST | N - 1 | - | -
Model Fit Metrics: R 2 R^2 R 2 , R adj 2 R^2_{\text{adj}} R adj 2 , and R pred 2 R^2_{\text{pred}} R pred 2
Ordinary R 2 R^2 R 2 : R 2 = 1 − S S Error S S Total R^2 = 1 - \frac{SS_{\text{Error}}}{SS_{\text{Total}}} R 2 = 1 − S S Total S S Error . Measures proportion of variation explained. (Always increases when factors are added).
Adjusted R 2 R^2 R 2 : R adj 2 = 1 − M S Error M S Total / d f Total R^2_{\text{adj}} = 1 - \frac{MS_{\text{Error}}}{MS_{\text{Total}} / df_{\text{Total}}} R adj 2 = 1 − M S Total / d f Total M S Error . Penalizes model complexity.
Predicted R 2 R^2 R 2 : R pred 2 = 1 − P R E S S S S Total R^2_{\text{pred}} = 1 - \frac{PRESS}{SS_{\text{Total}}} R pred 2 = 1 − S S Total PRESS . Measures predictive capability on new data using Prediction Error Sum of Squares ( P R E S S PRESS PRESS ).
3. Response Surface Methodology (RSM)
When factor screening reveals non-linear behavior (curvature) in the response, Response Surface Methodology (RSM) is employed to model second-order quadratic surfaces and locate optimal process operating conditions.
Second-Order Response Surface Model y = β 0 + ∑ i = 1 k β i x i + ∑ i = 1 k β i i x i 2 + ∑ i < j β i j x i x j + ϵ y = \beta_0 + \sum_{i=1}^k \beta_i x_i + \sum_{i=1}^k \beta_{ii} x_i^2 + \sum_{i<j} \beta_{ij} x_i x_j + \epsilon y = β 0 + i = 1 ∑ k β i x i + i = 1 ∑ k β ii x i 2 + i < j ∑ β ij x i x j + ϵ
Where β i i \beta_{ii} β ii represents pure quadratic curvature terms.
Sequential Optimization Strategy
Screening Phase : Fractional Factorial ( 2 k − p 2^{k-p} 2 k − p ) to eliminate inactive factors.
Steepest Ascent Phase : Follow the gradient of maximum response change to move toward the optimal region.
Response Surface Phase : Fit a full quadratic model using a second-order design ( Central Composite or Box-Behnken ).
4. Central Composite Design (CCD) & Box-Behnken Design (BBD)
Fitting a full second-order model requires factors to be evaluated at a minimum of 3 3 3 levels. The two principal RSM designs are CCD and BBD.
Central Composite Design (CCD)
A CCD is constructed by augmenting a standard 2 k 2^k 2 k factorial (or 2 k − p 2^{k-p} 2 k − p Resolution V) design with axial (star) points and center points:
2 k 2^k 2 k Factorial Points : Code values ( ± 1 , ± 1 , … , ± 1 ) (\pm 1, \pm 1, \dots, \pm 1) ( ± 1 , ± 1 , … , ± 1 ) . Estimates linear and 2-factor interaction terms.
2 k 2k 2 k Axial (Star) Points : Code values ( ± α , 0 , … , 0 ) , ( 0 , ± α , … , 0 ) (\pm \alpha, 0, \dots, 0), (0, \pm \alpha, \dots, 0) ( ± α , 0 , … , 0 ) , ( 0 , ± α , … , 0 ) , etc. Estimates pure quadratic terms β i i \beta_{ii} β ii .
n c n_c n c Center Points : Code value ( 0 , 0 , … , 0 ) (0, 0, \dots, 0) ( 0 , 0 , … , 0 ) . Provides pure error estimation and tests for overall curvature.
Total CCD Runs N = 2 k + 2 k + n c N = 2^k + 2k + n_c N = 2 k + 2 k + n c . Central Composite Design (CCD) for k=2
- (-1, 1) (1, 1)
- o---------------o
- | |
- (-α, 0) o | (0,0) o | o (α, 0) <-- Axial Points
- | |
- o---------------o
- (-1,-1) (1,-1)
- Rotatability Condition in CCD
A design is rotatable if the variance of the predicted response y ^ \hat{y} y ^ is identical at all points equidistant from the design center. For a CCD, rotatability is achieved by choosing the axial distance α \alpha α : α = ( N F ) 1 / 4 \alpha = (N_F)^{1/4} α = ( N F ) 1/4
Where N F = 2 k N_F = 2^k N F = 2 k is the number of factorial points.
For k = 2 k = 2 k = 2 factors: N F = 4 ⟹ α = ( 4 ) 1 / 4 = 2 ≈ 1.414 N_F = 4 \implies \alpha = (4)^{1/4} = \sqrt{2} \approx 1.414 N F = 4 ⟹ α = ( 4 ) 1/4 = 2 ≈ 1.414 .
For k = 3 k = 3 k = 3 factors: N F = 8 ⟹ α = ( 8 ) 1 / 4 ≈ 1.682 N_F = 8 \implies \alpha = (8)^{1/4} \approx 1.682 N F = 8 ⟹ α = ( 8 ) 1/4 ≈ 1.682 .
Box-Behnken Design (BBD)
A Box-Behnken Design is an independent, spherical 3-level design formed by combining 2 2 2^2 2 2 factorials with central block points. Unlike CCD, BBD does not contain corner points (where all factors are simultaneously at extreme levels ± 1 \pm 1 ± 1 ) or extreme axial points ( ± α \pm \alpha ± α ).
Design Feature Central Composite Design (CCD) Box-Behnken Design (BBD)
Number of Levels 5 levels ( − α , − 1 , 0 , + 1 , + α ) (-\alpha, -1, 0, +1, +\alpha) ( − α , − 1 , 0 , + 1 , + α ) 3 levels ( − 1 , 0 , + 1 ) (-1, 0, +1) ( − 1 , 0 , + 1 )
Extreme Combinations Contains extreme corner and star points. Excludes extreme corners; safer when extremes cause process failure.
Efficiency ( k = 3 k=3 k = 3 ) 15 − 20 15 - 20 15 − 20 runs required. 13 − 15 13 - 15 13 − 15 runs required.
Sequentiality Can build upon an existing 2 k 2^k 2 k factorial design. Cannot be built sequentially from a 2 k 2^k 2 k factorial design.