| |
:: Course SummaryThis 1-day training reviews how to construct advanced experimental designs to screen for the most influential factors to study in future experiments or studies. These designs require very few runs, thus minimizing the time and cost needed at this preliminary phase.:: Learning ObjectivesUpon completion of this training course, participants will be able to:
Understand the limitations of full factorial designs when the number of factors to study is very large
Choose an appropriate screening design for their type of studies and objectives
Construct the design selected
Analyze the results of the study or experiment
Interpret the results of screening designs
Differentiate between and critique the results of different screening designs, such as the Taguchi design:: Target AudienceThis training is primarily intended for people conducting preliminary studies or experiments, who wish to understand the impact of studying the effect of a large number of factors on a given phenomenon and who require tools in selecting the most influential factors to study in later phases of the experimental process. This course is also intended for researchers seeking to reduce the number of factors to study in an experiment.:: PrerequisiteParticipants must have a working knowledge of the Analysis of Variance (ANOVA) method as well as the construction of simple factorial designs, or, equivalently, have followed the courses:
Fundamental Tools in Statistics
Introduction to the Design of Experiments | | |
:: Topics Covered
- Why Design Experiments
- Strengths and Weaknesses of Complete Factorial Designs
- Principles of Fractional Designs (Screening Designs)
- Reducing the Number of Runs
- The Notion of Modeling in Fractional Designs
- Integrating the Study Objective in the Experimental Design
- Impact of Fractioning
- The Role of Repetition and Balance in Designs
- Constructing Screening Designs
- Underlying Assumptions
- Basics of Construction
- A Simple Case: A Design with 3 Factors at 2 Levels Each
- Generalizing to Higher Order Designs: Resolution of the Design, Choice of Screening Design and Construction of the Design
- Analyzing Study Results
- Different Types of Screening Designs
- Plackett-Burman Designs
- Designs for 3-Level Factors
- Mixed Designs
- Advanced Designs
- Overview of Taguchi Designs
- Principles
- Examples
- Critique of Taguchi Designs
- Summary
:: Course ContentThe number of factors that can have an influence on an outcome of interest usually come in great numbers. Whether controllable or uncontrollable factors, each of these factors can play a role in the variability observed in a measured phenomenon. At a preliminary experimentation stage, it is crucial to use appropriate statistical techniques to sift through all these factors and retain the most influential ones.
Under full factorial designs, where all possible factor combinations are tested, the number of runs needed to determine which factors to retain quickly becomes infeasible. Due to time and cost constraints, special designs are required to test all factors of interest while keeping the number of runs at a reasonable level.
Screening designs, as the name implies, are specially conceived to address this research problem. Also called fractional designs, these designs can test the essential effects to test while using only a fraction of the total number of runs under a full factorial design.
This training session reviews the different screening designs available, how to select the most appropriate one in different research settings, how to construct the design selected and determine the number of runs required and how to analyze, interpret and understand the scope of the results from such designs.
The screening designs discussed include fractional designs such as the Plackett-Burman design, designs for factors with three levels, mixed designs and Taguchi designs. Examples and case studies will be presented to provide direct applications of these designs. For each design, its use, construction method, strengths and weaknesses will be presented.
| |
