Quality and Productivity Improvement Using the SCA Statistical System

Products and processes must be designed for quality, right from the start. To determine which factors are central to improvement, critical events (those containing significant information) and perceptive observers (those who can exploit the information) must be brought together. Active scientific intervention is the most efficient means to accomplish this task. The SCA Quality Improvement Package provides a bridge to bring critical events and perceptive observers together.

The experimental design capabilities of the SCA Quality Improvement Package are keyed to the popular texts, Statistics for Experimenters by Box, Hunter and Hunter (1978), and Empirical Model-Building and Response Surfaces by Box and Draper (1987).

Powerful and Comprehensive
One aspect of the capabilities of the SCA Quality Improvement Package is the use of two-level factorial designs. This contemporary approach allows you to study the effects of more than one variable simultaneously. Since factors often interact, two-level full and fractional factorial make the most efficient use of resources for experimentation.

The SCA Quality Improvement Package provides a variety of features, from simple plots and descriptive statistics to more advanced features. The capabilities of the SCA Quality Improvement Package are contained within two modules (SCA-QPI and SCA-GSA on workstations and mainframes, or PC-QPI and PC-GSA on personal computers). There is no difference in capability between the mainframe, workstation and personal computer versions of the SCA Quality Improvement Package.


SCA-QPI
  • System-aided contrunction of two-level full and fractional designs
  • Complete analyses of two-level designs
  • Response surface analysis
  • Analysis for balanced and unbalanced factorial designs
  • Product design insensitive to environmental factors
  • Robust product designs
  • New aspects of Box-Cox transformations
  • Advanced regression capabilities
  • Control charts

Creating two-level designs

The DESIGN feature of the SCA-QPI module guides you through a sequence for the construction of two-level designs. This opens for the experimenter:

  • Replication of full and fractional factorial designs
  • Blocking and randomization
  • Placket-Burman designs
  • Planning for future experiments by considering
  • Foldover and switch column designs
  • Adding and dropping factors

Whenever a design is created or altered, the alias structure of the design is displayed. This provides you with valuable information about what effects and interactions will be confounded with one another in the subsequent analyses of the design. This knowledge will help you choose a design most appropriate for the problem at hand. The DESIGN feature can automatically produce a design of highest resolution.

Complete analysis of experimental designs

The SCA-QPI module makes experimental results readily understandable. Graphical displays highlight results. Powerful and accurate algorithms provide useful statistics and relevent information about the experiment. Informational output provided by the SCA-QPI module includes:

  • Cube plots (to observe responses at various factor level combinations)
  • Dispersion plots (to observe variation of response at various factor combinations)
  • Calculation and display of estimated effects
  • Alias structure for estimates
  • Probability plots (to visually determine active, or real, effects)
  • Bayes plots (to analytically determine and display active effects)
  • Analysis for balanced and unbalanced designs (for one-way to N-way factorials)
  • Analysis of variance and covariance (for one-way to N-way factorials)

New aspects of Box-Cox transformations

Frequently one or more of the assumptions of a statistical model are not satisfied by data. This results in inefficient and unnecessarily complicated analyses. Box and Cox showed how a power transformation of the response can improve an analysis and reveal useful information. More recently, Box and Fung have employed graphical displays, lambda plots, to show how an appropriate transformation can simultaneously simplify a model and also yield more efficient parameter estimates. This new analysis is featured in the SCA System.

Product design insensitive to environmental factors

An important characteristic ofa quality product is that it performs well over the actual range of environmental conditions it may encounter. Statistical measures can be employed to determine a product's performance over oeprating conditions, as well as identifying factors affecting performance. The SCA-QPI module provides valuable capabilities in this regard. Among them are:

  • Robust product design
  • Various performance criteria
  • Signal-to-noise ratios
  • Coefficient of variation
  • Weighted mean and weighted loss function
  • Mean and standard deviation of original or logged data

Response surface analysis

Response surface analysis helps an investigator find a combination of factors which may yield an optimum response. It can also be used to enable a better product design while simultaneously satisfying desired specifications. The SCA-QPI product provides you with capabilities to:

  • Analyze a response design of any polynomial order
  • Conduct a ridge analysis of surfaces
  • Conduct a Box-Cox transformation analysis
  • Implement evolutionary operation (EVOP) techniques

EVOP is a means for both the continuos operation and monitoring of a process. It is often regarded as an important managerial tool in which investigative methods become part of basic operations. In this manner, a product can be produced within tolerance limits while information is gathered to help improve the process.

Surveillance methods

Active scientific experimentation is important to improve the quality of a product or process. Also important is the surveillance of the quality of the product or process. Surveillance methods, such as control charts and sampling inspection, can be used to indicate changes in quality or to detect causes of quality variation. The SCA-QPI module provides an array of surveillance tools such as:

  • Pareto diagrams
  • Shewart control charts (Xbar-R and Xbar-S)
  • p-charts and np-charts
  • c-charts and u-charts


SCA-GSA

The SCA-GSA module provides a wide range of general statistical capabilities from graphical to selected advanced analysis features. As a component of the SCA Quality Improvement Package, the SCA-GSA module provides a more complete range of features for experimental design and statistical analysis. With these integrated capabilities, the SCA Quality Improvement Package extends the scope of its applications beyond experimental design and analysis. With SCA-GSA, you also have access to:

  • Plots and descriptive statistics
  • Regression analysis, including serially correlated errors
  • Analysis of variance
  • Cross tabulation
  • Box-Jenkins ARIMA modeling
  • Nonparametric statistics
  • Contingency tables and chi-square tests
  • Analytic functions and matrix operators
  • and more ...