Statistical Methods in Lean Management, Six Sigma, Continuous improvement Introduction

This curriculum spans the statistical rigor of a multi-workshop Six Sigma Black Belt program, integrating foundational to advanced methods as applied in cross-functional process improvement initiatives across manufacturing and transactional environments.

Module 1: Foundations of Statistical Thinking in Process Improvement

• Selecting between descriptive and inferential statistics based on data availability and project phase in a manufacturing environment.
• Defining operational definitions for critical-to-quality (CTQ) metrics to ensure consistent data collection across shifts and departments.
• Choosing appropriate data types (continuous vs. discrete) during measurement system analysis to align with process capability requirements.
• Implementing stratified sampling strategies when production batches exhibit known variation by machine or operator.
• Validating data normality using graphical (Q-Q plots) and statistical (Anderson-Darling) methods before applying parametric tests.
• Documenting assumptions and limitations of baseline performance metrics for audit and regulatory compliance in regulated industries.

Module 2: Measurement System Analysis and Data Integrity

• Designing Gage R&R studies with cross-functional team input to reflect real-world operator variation in assembly processes.
• Setting acceptance criteria for %GRR based on process tolerance and criticality, balancing cost of measurement error against rework risk.
• Deciding between attribute and variable MSA based on inspection method feasibility and engineering specifications.
• Integrating calibration schedules with MSA results to maintain measurement reliability in high-volume production lines.
• Addressing non-replicable measurements (e.g., destructive testing) using nested ANOVA models and specialized sampling plans.
• Establishing escalation protocols when MSA reveals unacceptable reproducibility across multiple shifts.

Module 3: Process Capability and Performance Analysis

• Selecting between Cp/Cpk and Pp/Ppk based on data collection time frame and process stability verification.
• Adjusting capability indices for non-normal data using transformations (e.g., Box-Cox) or non-parametric methods (e.g., percentiles).
• Handling short-run processes by applying group tolerance charts or Z-score normalization across product families.
• Setting realistic capability targets that align with customer specifications and current process technology limits.
• Interpreting confidence intervals for Cpk to assess risk in supplier qualification decisions.
• Updating capability assessments after process changes, ensuring data reflects post-improvement stability.

Module 4: Control Charts and Statistical Process Control

• Choosing between Xbar-R, Xbar-S, I-MR, and attribute charts based on subgroup size and data type in transactional vs. production settings.
• Establishing rational subgroups by analyzing process flow and identifying natural cycles or shifts.
• Setting control limits using initial stable data, then freezing them for ongoing monitoring during improvement phases.
• Responding to out-of-control signals with documented investigation workflows to distinguish special cause from common cause variation.
• Implementing pre-control charts in startup phases where historical data is insufficient for traditional SPC.
• Integrating control chart outputs with automated process shutdown systems in high-speed manufacturing environments.

Module 5: Hypothesis Testing for Process Comparisons

• Selecting between t-tests, ANOVA, and non-parametric alternatives (e.g., Mann-Whitney) based on data distribution and variance equality.
• Calculating required sample sizes using power analysis to detect meaningful process shifts without excessive data collection.
• Managing multiple comparisons in multi-line or multi-plant studies using Bonferroni or Tukey adjustments.
• Interpreting p-values in context of practical significance, especially when small differences are statistically significant but operationally irrelevant.
• Structuring paired tests for before-and-after comparisons when process changes cannot be rolled back.
• Documenting test assumptions and violations in project reports for regulatory or internal audit review.

Module 6: Design of Experiments (DOE) in Process Optimization

• Choosing between full factorial, fractional factorial, and response surface designs based on resource constraints and interaction effects of interest.
• Blocking experimental runs by shift or raw material lot to control for known sources of variation.
• Randomizing run order in constrained environments where equipment setup time affects feasibility.
• Handling hard-to-change factors using split-plot designs and appropriate error term selection.
• Validating model adequacy through residual analysis and lack-of-fit testing before drawing conclusions.
• Deploying confirmation runs under standard operating conditions to verify predicted improvements.

Module 7: Regression and Predictive Modeling for Continuous Improvement

• Selecting predictor variables using domain knowledge and correlation analysis to avoid overfitting in small datasets.
• Assessing multicollinearity among process inputs when building multiple regression models for yield prediction.
• Validating model assumptions (linearity, homoscedasticity, independence) using residual diagnostics in time-series process data.
• Deploying logistic regression for defect prediction when outcome is binary and inputs include both continuous and categorical factors.
• Updating regression models periodically to reflect process drift or equipment upgrades.
• Communicating prediction intervals to operations teams to set realistic expectations for model-based forecasts.

Module 8: Integration of Statistical Methods in Lean and Six Sigma Deployment

• Aligning statistical tool selection with DMAIC phase objectives to avoid premature hypothesis testing in Define.
• Standardizing data collection templates across Black Belt projects to ensure consistency in statistical reporting.
• Establishing governance thresholds for statistical significance in tollgate reviews to maintain methodological rigor.
• Coordinating statistical software access and version control across global teams to ensure reproducible analysis.
• Training Green Belts on correct interpretation of control charts and capability indices to reduce misapplication.
• Embedding statistical review checkpoints in project charters to prevent flawed data collection designs.