Examples
From the question through the choice of design, the design with its measurements, model and analysis of variance to the confirmed setting: four complete analyses as they come about in DoEStat. Every table and every chart comes out of the software — and every example ships with it as a project, to be run yourself.
01Screening
Warpage in injection moulding: which of ten machine settings matter? A minimum-run resolution IV design finds four factors and one interaction — and the centre runs show that the surface is curved.
02Response surface
An API synthesis without constraints: yield, by-product and residual palladium pull in different directions. A central composite design with 30 runs and desirability find the compromise.
03Mixture
An emulsion paint: lower and upper bounds, a sum constraint and a ratio constraint leave a body with 28 vertices of the mixture space. An I-optimal design with 25 batches leads to the best feasible recipe for three responses.
04Sequential
Tabletting in two campaigns: the six-factor screening ends with an optimum on the edge. The 19 runs are carried into a slightly shifted region of the three active factors and augmented by 15 — five responses, one window.
About the data. The measurements of all four examples are simulated: from models we set ourselves, plus random scatter. That is deliberate — only then can the end of each example show what the analysis found and what it did not. The models are given in each section “Cross-check”. Examples with published measurements from the literature ship with DoEStat as well.
30 days, the full feature set, no payment details. We send the download link by e-mail, usually on the next business day.
