Summa Six Sigma
This course covers a number of advanced statistical methods useful for individuals engaged in Six Sigma projects. The techniques described go beyond those usually presented in Six Sigma training programs. However, they deal with important situations that are often encountered when analyzing real data.

Module length: 2 days

Prerequisites: Attendees should be familiar with standard Six Sigma statistical methods, including capability analysis, control charts, regression analysis, and design of experiments. Knowledge of the material covered in the BASIC, SPC1 and DOE1 modules, described on the Course Descriptions page, is sufficient.

Outline:

Multivariate Capability Analysis

Multivariate Normal Distribution
Estimation of Joint Probability of Being Within Spec.
Multivariate Capability Indices

Multivariate Process Control

Hotelling's T-Squared
T-Squared Control Charts
Multivariate EWMA Control Charts
Generalized Variance Charts
Use of Principal Components Analysis or PLS with Control Ellipses

Control Charts and Capability Analysis for Nonnormal Data

Probability Distributions for Skewed Data
Probability Distributions for Data with Significant Kurtosis
Generalized Gamma, Generalized Logistic, Exponential Power Distributions
Selecting the Proper Distribution
Control Limits for Nonnormal Data
Capability Indices for Nonnormal  Data
Transformation Methods

Outlier Identification and Accommodation

Grubbs, Dixon's and Tukey's Tests
Accommodation Methods (Trimming, Winsorization)

Control Charts for Autocorrelated Data

Identifying and Estimating ARIMA Models
Modifying Control Limits
Residual Control Charts

Special Purpose Control Charts

Modifying Control Limits for High Cpk Processes
Cuscore Charts for Detecting Special Patterns
Toolwear Charts for Trending Data

Regression Analysis and Classification

Fitting Nonlinear Models
Discriminant Analysis
Bayesian Neural Networks

Multivariate Optimization

Multivariate Desirability Functions
Following the Path of Steepest Ascent

Automatic Forecasting

Forecasting Methods
Model Selection Criteria