Orthogonal Arrays of Strength 3 with Full Estimation Capacities
Orthogonal arrays of strength 3 permit estimation of all the main effects of the experimental factors free from contamination with 2-factor interactions (2fi). If they allow for estimation of all these interactions, they are said to have full estimation capacity (FullEC). A necessary condition for the existence of FullEC arrays of strength 3 is in the number of parameters not exceeding the run size. If such arrays exist, they are a more economical alternative to arrays of strength 4, at the price of a less efficent estimation of the interactions. So there might be FullEC designs with the same run-size allowing a more efficent estimation of these effects. We consider the practical choice between FullEC strength-3 arrays as compared to competitor designs that are more efficent regarding estimation of the 2fi We generate examples by enumeration of all non-isomorphic arrays of six 48-run series of mixed orthogonal arrays of strength 3. Competitor arrays were generated using a modifid Fedorov algorithm for D-optimum designs. We evaluate efficencies of the arrays under models with all the main effects and all the interactions save those of k of the factors, with k∈0, 1,2 . Of particular use are efficencies for estimation of the main effects with the 2fi as nuisance parameters, and for estimation of these interactions with the main effects as nuisance parameters. If a few of the factors might be inactive, while their identity is not known in advance, we recommend the use of specific cases selected from the series of strength-3 arrays.
Experimental Design Non-isomorphic Designs Mixed Array Ds-optimum Designs.
Eric D. SCHOEN Man V.M. NGUYEN
Dept. of Mathematics and Computer Science Eindhoven University of Technology Eindhoven, Holland Division of Computational Science University of Technology Ho Chi Minh City, Vietnam
国际会议
2006 International Conference on Design of Experiments and Its Applications(2006实验设计及其应用国际会议)
天津
英文
2006-07-09（万方平台首次上网日期，不代表论文的发表时间）