Using Blocking When Designing Experiments

See how to:

• Understand the restrictions associated with randomization, the difference between blocking and creating plots, and how to incorporate covariates into the design (1:26)
  • Blocks are about variance and let you group homogeneous factors to isolate extraneous noise associated with the factors (like time of day)
  • Plots address variability and let you organize units so you can estimate the noise/signal associated with factors
  • Incorporating covariates (uncontrolled factors related to the experiment and known before the experiment) lets you asses the impact of the related factors
• Understand the factors related to the widget case study where manufacturers add hardening agent to slurry, stir it, control viscosity, mold it and then bake it (7:06)
• Use Custom Designer to incorporate differences related to differences related to the day the experiment runs (8:55)
  • See how to group runs into blocks to estimate variability from day to day (Random Effect blocks)
  • See how to add Blocking Factor to calculate mean differences from day to day (Fixed Effect blocks)
  • Understand the impact of choosing Random vs. Fixed Effects on the number of runs needed for the design.
  • Understand why to choose a Split-Plot design to get estimates of signal plus noise
• Use Custom Designer to build Split-Plot design that includes statistics related to interactions between hard-to-change and all other factors (18:28)
• Compare incorrect and correct designs (27:00)
  • Examine Fixed Effect Test results
• Understand and compare Split-Split-Plot and two-way Split-Plot (Strip Plot) Designs (30:16)
• Add information about a related factor (covariate) to the model (38:43)
  • Handle mis-match between number of covariate observations and number of runs needed

Note: Includes Q& A at times 36:30, 43:53. 44:49 and 45:16.

Resources

• Optimal Design of Experiments: A Case Study Approach (Goos and Jones) ISBN: 978-1-119-97616-5, June 2011, 304 Pages
• Design Diagnostics video (McCormack). Covers Custom Design, Definitive Screening Design and an augmented Plackett-Burman-like design for a product with two categorical and seven continuous factors.
• Advanced Design of Experiments video (McCormack). Covers working with linear constraints and handling blocking and hard-to-change factors.
• Split-Plot and Strip-Plot Design of Experiments video (McCormack)
• DOE Case Studies with Professor Peter Goos videos. Covers split-plot designs.