22: Full Factorial Designs
"Block what you can, randomize what you cannot." - Unknown. "Mathematicians do not study objects, but relations among objects." — Henri PoincaréLecture outline: how to conduct experiments in which multiple factors are changed simultaneously?
Preamble: interaction between factors
1. Two level factorial design without replications (2^k designs)
When to use two level designs?
2^2 design: model and example
Computation of effects: sign-table method
Allocation of variance
General 2^k factorial designs
Case study: 2^3 factorial design.
2. Two level factorial design with replications (2^k).r designs
Computation of effects
Estimation of experimental errors
Allocation of variation
Primary reference for this lecture:
“The Art of Computer Systems Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and Modeling” by Raj Jain; Selection sections from Chapters 21, 22, 23.