A Ridge-Type Procedure for Design of Experiments

Abstract

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika. Miscellanea 147When the centre point is somewhat more heavily weighted than the other points in the design, the procedure generally followed in practice, the design will perform well by all criteria. Box & Behnken recommended three centre points to make the variance profile relatively flat near the centre of the design. This increases the E-and A-efficiencies and decreases the D-and G-efficiencies slightly. As additional centre points also allow an independent estimate of experimental error, their recommendation is a good one.The centre point weight giving the highest E-efficiency has the lowest efficiencies by the other criteria. E-optimality can be described as minimizing the variance of the linear combination of regression coefficients having maximum variance. For a second or higher order polynomial model, this optimization, which combines regression coefficients of different orders, is less intuitively pleasing than the other optimality criteria. The other optimality criteria illustrate Kiefer's (1975) point that finding a design optimum by one criterion often gives a design that performs well by other criteria. REFERENCES

A Simple Bayesian Modification of D-Optimal Designs to Reduce Dependence on an Assumed Model

“…The posterior mean b = b(r) is interpreted as a Bayesian shrinkage estimator of the coefficients or merely as a ridge regression estimator. See Efron and Morris (1973), Oman (1984), and Vuchkov (1977) for discussion of such estimators. One can examine the ridge trace-namely, b(r) as r varies-to evaluate the sensitivity of the parameter estimates to the prior distribution.…”

On greedy heuristics for computing D-efficient saturated subsets

“…Except for the simple or the weighted random choice of a saturated set S ⊆ F, the most common is the procedure based on a regularization of the information matrix of a sub-saturated set (see, e.g., Section 11.2 of [2]; cf. [31]). The regularized greedy heuristic (RGH) specifies step (S) as…”

Optimal Design of Experiments on Riemannian Manifolds

“…As far as we know, the first discussion of regularized ODOE comes from Vuchkov (1977), who proposed a ridge-type procedure for ODOE based on the ridge regression estimator: βridge = argmin…”