Box-Behnken, Central-Composite, D and I-optimal designs were compared using statistical tools. Experimental trials for all designs were generated. Random uniform responses were simulated for all models. R-square, Akaike and Bayesian Information Criterion for the fitted models were noted. One-way ANOVA and Tukey's multiple comparison test were performed on these parameters. These models were evaluated based on the number of experimental trials generated in addition to the results of the statistical analyses. D-optimal design generated 12 trials in its model, which was lesser in comparison to both Central Composite and Box-Behnken designs. The R-square values of the fitted models were found to possess a statistically significant difference (P<0.0001). D-optimal design not only had the highest mean R-square value (0.7231), but also possessed the lowest means for both Akaike and Bayesian Information Criterion. The D-optimal design was recommended for generation of response surfaces, based on the assessment of the above parameters. © Published under licence by IOP Publishing Ltd.
View more info for "Selection of a design for response surface"
|Journal||Data powered by TypesetIOP Conference Series: Materials Science and Engineering|
|Publisher||Data powered by TypesetIOP Publishing|