Worksheets/Week6

From Statistics for Engineering
Revision as of 09:00, 17 October 2019 by Kevin Dunn (talk | contribs) (→‎Part 1)
Jump to navigation Jump to search

Part 1

Case study: Achieve a stability value of 50 days or more, for a new product. We had a full factorial set of experiments in 3 factors:


  • B: feed concentration: -1 == 5%; +1 == 15%
  • C: mixer type: -1 = R mixer; +1 = W mixer
Factor name Description Low value High value Type of factor
A Enzyme strength 20% 30% Numeric factor


A: enzyme strength [numeric factor] B: feed concentration [numeric factor] C: mixer type [categorical factor] y = Stability [days]


In this notebook we will show what we loose out if we pretend we only did half the experiments. In other words, we actually have 8 experiments, but we will see what happens if we only use 4 of them.



# This is the half-fraction, when C = A*B A = c(-1, +1, -1, +1) B = c(-1, -1, +1, +1) C = A * B # The response: stability [units=days] y = ... model_stability_poshalf model_stability_poshalf = lm(y ~ A*B*C) summary(model_stability_poshalf) # Uncomment this line if you run the code in RStudio #library(pid) # Comment these 2 lines if you run this code in RStudio source('https://yint.org/paretoPlot.R') source('https://yint.org/contourPlot.R') paretoPlot(model_stability_poshalf) # This is the other half-fraction, when C = -A*B A = c(-1, +1, -1, +1) B = c(-1, -1, +1, +1) C = -1 * A * B # The response: stability [units=days] y = ... model_stability_neghalf = lm(y ~ A*B*C) summary(model_stability_neghalf)

Part 2

Data from a bioreactor experiment is available, were we were investigating four factors:

  • A = feed rate: 5 g/min or 8 g/min
  • B = initial inoculate amount: 300 g or 400 g
  • C = feed substrate concentration: 40 g/L or 60 g/L
  • D = dissolved oxygen set-point: 4 mg/L or 5 mg/L

The 16 experiments from a full factorial, 24, were randomly run, and the yields, y, the outcome variable were given in standard order: [60, 59, 63, 61, 69, 61, 94, 93, 56, 63, 70, 65, 44, 45, 78, 77]

base <- c(-1, +1) design <- expand.grid(A=base, B=base, C=base) A <- design$A B <- design$B C <- design$C D <- A * B * C # Confirm that you can find these 8 runs yourself: y <- c(60, 63, 70, 61, 44, 61, 94, 77) model.bio <- lm(y ~ A*B*C*D) summary(model.bio) # Uncomment this line if you run the code in RStudio #library(pid) # Comment these 2 lines if you run this code in RStudio source('https://yint.org/paretoPlot.R') source('https://yint.org/contourPlot.R') paretoPlot(model.bio)