Approximate D-optimal designs of experiments on the convex hull of a finite set of information matrices

In the paper we solve the problem of D _ℋ-optimal design on a discrete experimental domain, which is formally equivalent to maximizing determinant on the convex hull of a finite set of positive semidefinite matrices. The problem of D _ℋ-optimality covers many special design settings, e.g., the D-optimal experimental design for multivariate regression models. For D _ℋ-optimal designs we prove several theorems generalizing known properties of standard D-optimality. Moreover, we show that D _ℋ-optimal designs can be numerically computed using a multiplicative algorithm, for which we give a proof of convergence. We illustrate the results on the problem of D-optimal augmentation of independent regression trials for the quadratic model on a rectangular grid of points in the plane.     

Optimal designs for estimating the coefficients of the lower frequencies in trigonometric regression models

In the common Fourier regression model we determine the optimal designs for estimating the coefficients corresponding to the lower frequencies. An analytical solution is provided which is found by an alternative characterization of c-optimal designs. Several examples are provided and the performance of the D-optimal design with respect to the estimation of the lower order coefficients is investigated. The results give a complete answer to an open question which was recently raised in the literature.     

D-Optimal designs for weighted polynomial regression—A functional approach

This paper is concerned with the problem of computing approximateD-optimal design for polynomial regression with analytic weight function on a interval [m ₀-a,m ₀+a]. It is shown that the structure of the optimal design depends ona and weight function. Moreover, the optimal support points and weights are analytic functions ofa ata=0. We make use of a Taylor expansion to provide a recursive procedure for calculating theD-optimal designs.     

Optimal design for smoothing splines

In the common nonparametric regression model we consider the problem of constructing optimal designs, if the unknown curve is estimated by a smoothing spline. A special basis for the space of natural splines is introduced and the local minimax property for these splines is used to derive two optimality criteria for the construction of optimal designs. The first criterion determines the design for a most precise estimation of the coefficients in the spline representation and corresponds to D-optimality, while the second criterion is the G-optimality criterion and corresponds to an accurate prediction of the curve. Several properties of the optimal designs are derived. In general, D- and G-optimal designs are not equivalent. Optimal designs are determined numerically and compared with the uniform design.     

Optimal and efficient designs for Gompertz regression models

Gompertz functions have been widely used in characterizing biological growth curves. In this paper we consider D-optimal designs for Gompertz regression models. For homoscedastic Gompertz regression models with two or three parameters, we prove that D-optimal designs are minimally supported. Considering that minimally supported designs might not be applicable in practice, alternative designs are proposed. Using the D-optimal designs as benchmark designs, these alternative designs are found to be efficient in general.     

Local c- and E-optimal Designs for Exponential Regression Models

In this paper we investigate local E- and c-optimal designs for exponential regression models of the form . We establish a numerical method for the construction of efficient and local optimal designs, which is based on two results. First, we consider for fixed k the limit μ _i → γ (i = 1, ... , k) and show that the optimal designs converge weakly to the optimal designs in a heteroscedastic polynomial regression model. It is then demonstrated that in this model the optimal designs can be easily determined by standard numerical software. Secondly, it is proved that the support points and weights of the local optimal designs in the exponential regression model are analytic functions of the nonlinear parameters μ ₁, ... , μ _k . This result is used for the numerical calculation of the local E-optimal designs by means of a Taylor expansion for any vector (μ ₁, ... , μ _k ). It is also demonstrated that in the models under consideration E-optimal designs are usually more efficient for estimating individual parameters than D-optimal designs.     

D-optimality andD n -optimal designs for mixtures regression models with logarithmic terms

In this paper the authors put forward a new mixtures regression models with Logarithmic terms^[14] to generalize Draper's models^{[1, 2 or 6]} by using Kiefer-Wolfowitz's equivalence theorem^[3], Fedorov's and Wynn's method^[5]. And we also suggest a method for computer-aided design of combinatorial search^[13].In this study, we have proved and constructed the approximateD-optimal (measure) andD _n -optimal (exact) designs by the use of the first and second order mixtures regression models with logarithmic terms in three and four components.Projects Supported by the Science Fund of the Chinese Academy of Sciences.     

On exact D-optimal designs for regression models with correlated observations

Let ^* be an exact D-optimal design for a given regression model Y = X + Z . In this paper sufficient conditions are given for sesigning how the covariance matrix of Z may be changed so that not only ^* remains D-optimal but also that the best linear unbiased estimator (BLUE) of stays fixed for the design ^*, although the covariance matrix of Z * is changed. Hence under these conditions a best, according to D-optimality, BLUE of is known for the model with the changed covariance matrix. The results may also be considered as determination of exact D-optimal designs for regression models with special correlated observations where the covariance matrices are not fully known. Various examples are given, especially for regression with intercept term, polynomial regression, and straight-line regression. A real example in electrocardiography is treated shortly.     

On the regularization of singular <Emphasis Type="Italic">c</Emphasis>-optimal designs

We consider the design of c-optimal experiments for the estimation of a scalar function h(θ) of the parameters θ in a nonlinear regression model. A c-optimal design ξ* may be singular, and we derive conditions ensuring the asymptotic normality of the Least-Squares estimator of h(θ) for a singular design over a finite space. As illustrated by an example, the singular designs for which asymptotic normality holds typically depend on the unknown true value of θ, which makes singular c-optimal designs of no practical use in nonlinear situations. Some simple alternatives are then suggested for constructing nonsingular designs that approach a c-optimal design under some conditions.     

Some optimal designs for comparing a set of test treatments with a set of controls

In this paper, we give a detailed study of the problem of optimally comparing a set of t test treatments to a set of s controls under a 0-way elimination of heterogeneity model. The relationships between designs that are A and MV-optimal for comparing the test treatments to the controls and those that are A and MV-optimal for comparing all treatments are also studied.Research is sponsored by NSF Grant No. DMS-8700945.     

A construction of D-optimal designs for N≡2 mod 4

D-optimal design of order 6 is used to construct D-optimal designs of order 42 and 66.     

Canonical forms of some special matrices useful in statistics

In experimental situations wheren two or three level factors are involved andn observations are taken, then theD-optimal first order saturated design is ann ×n matrix with elements ±1 or 0, ±1 with the maximum determinant. Canonical forms are useful for the specification of the non-isomorphicD-optimal designs. In this paper, we study canonical forms such as the Smith normal form, the first, second and the Jordan canonical form ofD-optimal designs. Numerical algorithms for the computation of these forms are described and some numerical examples are also given.     

Equilibrium probability calculations for a discrete-time bulk queue model

Many problems in management science and telecommunications can be solved by the analysis of aD ^X/D^m/1 queueing model. In this paper, we use the zeros, both inside and outside the unit circle, of the denominator of the generating function of the model to obtain an explicit closed-form solution for the equilibrium probabilities of the number of customers in the system. The moments of the number of customers in the queue or in the system are also studied. When there are infinitely many zeros outside the unit circle, we propose an approximation method using polynomials. This method yields correct values for a finite number of the probabilities, the number depending on the degree of the polynomial approximation.     

On the usage of refined linear models for determiningN-way classification designs which are optimal for comparing test treatments with a standard treatment

Summary In this paper we consider experimental settings in whichv test treatments are to be compared to some control or standard treatment and where heterogeneity needs to be eliminated inn-directions. Using techniques similar to those used by Kunnert (1983,Ann. Statist.,11, 247–257) concerning the determination of optimal designs under a refined linear model, some methods are given for constructingn-way classification designs which areA- andMV-optimal for estimating elementary treatment differences involving the standard treatment fromm-way classification designs,m<n, which areA- andMV-optimal for estimating the same treatment differences. Examples are given for the casen=2 to show how the results obtained can be applied. This research was supported by NSF grant No. DMS-8401943.     

Spherical containment and the Minkowski dimension of partial orders

The recent work on circle orders generalizes to higher dimensional spheres. As spherical containment is equivalent to causal precedence in Minkowski space, we define the Minkowski dimension of a poset to be the dimension of the minimal Minkowski space into which the poset can be embedded; this isd if the poset can be represented by containment with spheresS ^d–2 and of no lower dimension. Comparing this dimension with the standard dimension of partial orders we prove that they are identical in dimension two but not in higher dimensions, while their irreducible configurations are the same in dimensions two and three. Moreover, we show that there are irreducible configurations for arbitrarily large Minkowski dimension, thus providing a lower bound for the Minkowski dimension of partial orders.     

-Optimality conditions for composed convex optimization problems

The aim of the present paper is to provide a formula for the -subdifferential of f+gh different from the ones which can be found in the existent literature. Further we equivalently characterize this formula by using a so-called closedness type regularity condition expressed by means of the epigraphs of the conjugates of the functions involved. Even more, using the -subdifferential formula we are able to derive necessary and sufficient conditions for the -optimal solutions of composed convex optimization problems.     

ε-Value Function and Dynamic Programming

In this note, we develop a dynamic programming approach for an ε-optimal control problem of Bolza. We prove that each Lipschitz continuous function satisfying the Hamilton-Jacobi inequality (less than zero and greater than −ε) is an ε-value function.     

Duality for <Emphasis Type="Italic">ε</Emphasis>-Variational Inequality

In this paper, following the method in the proof of the composition duality principle due to Robinson and using some basic properties of the ε-subdifferential and the conjugate function of a convex function, we establish duality results for an ε-variational inequality problem. Then, we give Fenchel duality results for the ε-optimal solution of an unconstrained convex optimization problem. Moreover, we present an example to illustrate our Fenchel duality results for the ε-optimal solutions. The authors thank the referees for valuable suggestions and comments. This work was supported by Grant No. R01-2003-000-10825-0 from the Basic Research Program of KOSEF.     

Designs of ϕ-optimal control for second-order processes

Consider a realization of the process on the intervalT=[0,1] for functionsf ₁(t),f ₂(t),...,f _n (t) inH(R), the reproducing kernel Hilbert space with reproducing kernelR(s,t) onT×T, whereR(s,t)=E[ξ(s)ξt)] is assumed to be continuous and known. Problems of the selection of functions {f _k (t)} _k=1 ⁿ to be ϕ-optimal design are given, and an unified approach to the solutions ofD-,A-,E- andD _s-optimal design problems are discussed.