In preliminary research phases, the number of potentially influential factors to investigate is usually quite large. It is not uncommon to have to deal with at least 10 factors limiting the use of classic factorial designs.
Screening designs are experimental designs used to identify the most influential factors that influence a response or outcome in a process or system with a reasonable number of runs. These designs are typically used in the early stages of experimentation, when you want to quickly assess a large number of variables to determine which ones have the greatest effect on the response variable. The goal is to eliminate unimportant factors and focus resources on the most significant ones.
This course covers the construction of such designs, the price to pay to limit the number of factor level combinations to test and the statistical analysis of such data. The use of statistical models paramount in this course.
Participants must master ANOVA and linear regression modelling to attend this course.
Course Outline
The recommended duration for this course is 2 online session(s).
Session 1: Pros & Cons of Factorial Designs - The Idea Behind Factorial Fractional Designs- Estimable effects vs. experimentation cost
- How to reduce the number of runs?
- Notion of mathematical model
- Cost: Confounding & aliasing
- Notions of repetition & balance
- Basic designs
- Construction of 2k-p designs
- Selection criterion: design resolution
- Statistical analysis
- Plackett-Burman designs
- Use of Blocks in screening designs