A generalisation of T‐optimality for discriminating between competing models with an application to pharmacokinetic studies

The T‐optimality criterion is used in optimal design to derive designs for model selection. To set up the method, it is required that one of the models is considered to be true. We term this local T‐optimality. In this work, we propose a generalisation of T‐optimality (termed robust T‐optimality) that relaxes the requirement that one of the candidate models is set as true. We then show an application to a nonlinear mixed effects model with two candidate non‐nested models and combine robust T‐optimality with robust D‐optimality. Optimal design under local T‐optimality was found to provide adequate power when the a priori assumed true model was the true model but poor power if the a priori assumed true model was not the true model. The robust T‐optimality method provided adequate power irrespective of which model was true. The robust T‐optimality method appears to have useful properties for nonlinear models, where both the parameter values and model structure are required to be known a priori, and the most likely model that would be applied to any new experiment is not known with certainty. Copyright © 2012 John Wiley & Sons, Ltd.     

From Optimality Robustness To Sufficiency and Completeness

Let (\Omega,{\cal A},{\cal P}) stand for a statistical experiment and {\cal B},{\cal C} for some sub- σ -algebras of {\cal A} with {\cal C}\subset {\cal B} . It is shown that for any {\cal B} -measurable d\in\bigcap_{P\in {\cal P}}\,{\cal L}_{2}(\Omega,{\cal A},P) there exists some d_{1}\in\bigcap_{P\in {\cal P}}{\cal L}_{2}(\Omega,{\cal A},P) being {\cal C} -measurable and a UMVU estimator in (\Omega,{\cal A},{\cal P}) and some conditional white noise d_{2}\in\bigcap_{P\in {\cal P}}\,{\cal L}_{2}(\Omega,{\cal A},P) , i.e. E_{P}(d_{2}\vert {\cal C})=0,P\in {\cal P} , satisfying d=d_{1}+d_{2} , where d_{j},j=1,2 , are uniquely determined up to P -zero sets, if and only if {\cal C} is sufficient and complete for {\cal P}\vert {\cal B} and {\cal B} is optimality robust for {\cal P} , i.e. any {\cal B} -measurable d\in\bigcap_{P\in {\cal P}}\,{\cal L}_{2}(\Omega,{\cal A},P) being some UMVU estimator in the restricted statistical experiment (\Omega,{\cal B},{\cal P}\vert {\cal B}) is already a UMVU estimator in the original statistical experiment (\Omega,{\cal A},{\cal P}) . In particular, the special case {\cal B}={\cal A} characterizes sufficiency and completeness of {\cal C} for {\cal P} and the special case {\cal B}={\cal C} optimality robustness and completeness of {\cal C} for {\cal P} from a decomposition theoretical point of view. As an application it is shown that a σ -algebra containing a sufficient and complete sub- σ -algebra is optimality robust without being itself in general neither sufficient nor complete.     

Asymptotic Optimality of Sparse Linear Discriminant Analysis with Arbitrary Number of Classes

Many sparse linear discriminant analysis (LDA) methods have been proposed to overcome the major problems of the classic LDA in high‐dimensional settings. However, the asymptotic optimality results are limited to the case with only two classes. When there are more than two classes, the classification boundary is complicated and no explicit formulas for the classification errors exist. We consider the asymptotic optimality in the high‐dimensional settings for a large family of linear classification rules with arbitrary number of classes. Our main theorem provides easy‐to‐check criteria for the asymptotic optimality of a general classification rule in this family as dimensionality and sample size both go to infinity and the number of classes is arbitrary. We establish the corresponding convergence rates. The general theory is applied to the classic LDA and the extensions of two recently proposed sparse LDA methods to obtain the asymptotic optimality.     

On the choice of optimality criteria in comparing statistical designs

We formulate in a reasonable sense a class of optimality functionals for comparing feasible statistical designs available in a given setup. It is desired that the optimality functionals reflect symmetric measures of the lack of information contained in the designs being compared. In view of this, Kiefer's (1975) universal optimality criterion is seen to rest on stringent conditions, some of which can be relaxed while preserving optimality (in an extended sense) of the so-called balanced designs.     

Optimality of marginal likelihood estimating equations

In the presence of nuisance parameters, Godambe (1976) showed an optimality property of the conditional efficient score. We derive analogous results for marginal likelihood when the remainder of the likelihood is conditionally complete. The method of deriving this result is shown to be useful for deriving other optimality results. including optimality of the partial score.     

Simultaneous choice of time points and the block design in the growth curve model

The aim of this paper is to consider the optimality in the growth curve model with respect to two aspects: time and the block design and to show some relations between information functions for different designs. The A-, D- and E- optimality are studied.     

Some robust designs for polynomial regression models

The author identifies static optimal designs for polynomial regression models with or without intercept. His optimality criterion is an average between the D‐optimality criterion for the estimation of low‐degree terms and the D₈‐optimality criterion for testing the significance of higher degree terms. His work relies on classical results concerning canonical moments and the theory of continued fractions.     

Designs balanced for circular residual effects

Magda (1980) introduced a model for repeated measurements designs with a circular structure of the residual effects. He proved the universal optimality of circular balanced uniform designs over a subclass of the possible designs. We strengthen his result to optimality over the set of all designs with the same number of experimental units, periods and treatments.     

Optimal design for the proportional odds model

The authors construct locally optimal designs for the proportional odds model for ordinal data. While they investigate the standard D‐optimal design, they also investigate optimality criteria for the simultaneous estimation of multiple quantiles, namely D_A ‐optimality and the omnibus criterion. The design of experiments for the simultaneous estimation of multiple quantiles is important in both toxic and effective dose studies in medicine. As with c‐optimality in the binary response problem, the authors find that there are distinct phase changes when exploring extreme quantiles that require additional design points. The authors also investigate relative efficiencies of the criteria.     

On optimality of extremal schemes in progressive type II censoring

In a progressively type II censored life-testing experiment intact units may be removed from the experiment after every failure. If the initial number of units in the experiment and the total number of failures are fixed, the experimenter may choose between different censoring schemes. By specifying an optimality criterion, one may improve the outcome of the experiment by choosing the respective optimal scheme. We establish a simple property of a general optimality criterion that yields optimality of certain extremal schemes. Applications to some criteria that measure the total time of the experiment and its variability illustrate the approach. The results are based on stochastic orderings of generalized order statistics.     

Relationships of optimality for individual factors of a design

The optimality of two-factor experimental designs is studied in the dual senses of estimating contrasts in the parameters for each of the factors. The outline of comparison employed allows one to judge the performance of different designs for estimating contrasts of one set of parameters directly with the performance of the complementary set without going through a common intermediary step of considering all the parameters. The results hold for a wide class of optimality criteria (not merely D-, A- and E-optimality), which must satisfy a functional equation obtained in connection with our method. Also we investigate the optimality of row–column designs which satisfy an ‘adjusted orthogonality’ condition. Our point of departure is the paper by Shah, Raghavarao and Khatri (1976) and that of Mitchell and John (1977).     

Characterizing optimality among three-decision procedures for directional conclusions

Jones and Tukey (2000) proposed three-decision procedures for directional conclusions in statistical inference, considered as an alternative to the conventional usage of one- and two-tailed significance testing. Moreover, implicit in their suggestions was to consider a procedure to be optimal in case indefinite results were minimized among all procedures with a given control of error. First, we argue by example that this characterization of optimality is not very fruitful when formalized into the strong sense of uniform minimization. Next, imposing a further regularity condition on the comparative class of procedures, we relate the suggested characterization to optimality criteria from test theory (UMP unbiasedness). Similarly, we also consider characterizing optimality in terms of maximizing correct decision rates and minimizing incorrect decision rates. Finally, we demonstrate the applicability of the three considered characterizations with respect to exponential families of distributions.     

Optimal design of complex experiments with qualitative factors of influence

A complex experiment with qualirarive factors influencing the outcome of the experiment can be seen as a general ANOVA setup. A design of such an experiment will be the assignment at which of the possible levels of the factors the actual experiment should be performed. In this paper optimal designs of such experiments will be characterized with respect to three different optimality criteria including the so called uniform optimality of a design. The possible applications of the main optimization result providing these characterizations can be used to more general experiments. The particular results on these generalizations will be indicated at the end of this paper.     

Maximum likelihood mean and covariance matrix estimation constrained to general positive semi-definiteness

Maximum likelihood estimation of a mean and a covariance matrix whose structure is constrained only to general positive semi-definiteness is treated in this paper. Necessary and sufficient conditions for the local optimality of mean and covariance matrix estimates are given. Observations are assumed to be independent. When the observations are also assumed to be identically distributed, the optimality conditions are used to obtain the mean and covariance matrix solutions in closed form. For the nonidentically distributed observation case, a general numerical technique which integrates scoring and Newton's iterations to solve the optimality condition equations is presented, and convergence performance is examined.     

Balanced Asymmetrical Nearly Orthogonal Designs for first and second order effect estimation

A method for constructing asymmetrical (mixed-level) designs, satisfying the balancing and interaction estimability requirements with a number of runs as small as possible, is proposed in this paper. The method, based on a heuristic procedure, uses a new optimality criterion formulated here. The proposed method demonstrates efficiency in terms of searching time and optimality of the attained designs. A complete collection of such asymmetrical designs with two- and three-level factors is available. A technological application is also presented.     

ON AN APPROXIMATE OPTIMALITY CRITERION FOR THE DESIGN OF FIELD EXPERIMENTS UNDER SPATIAL DEPENDENCE

The design of large‐scale field trials where the residuals are correlated has been of recent interest, in large part because of advances in statistical and computational methods of analysis. The construction of designs for correlated data has typically used A‐optimality and is computationally intensive. This involves calculating the inverse of the information matrix for treatments under the supervision of an optimization strategy that explores the design space. We propose an approximation to A‐optimality, using nearest‐neighbour balance, that is less computationally demanding and can achieve at least 95% efficiency relative to A‐optimality in many practical situations.     

Optimal variance estimation for generalized regression predictor

The generalized regression (greg) predictor for the finite population total of a real variable is often employed when values of an auxiliary variable are available. Several variance estimators for it do well in large samples though bearing no optimality properties. We find a variance estimator which, under a restrictive model, has an optimality property under ‘exact’ as well as ‘asymptotic’ analysis. But this involves model parameters. Under a further restriction on the model, two model-parameter-free variance estimators are derived sharing the same ‘asymptotic’ optimality. Numerical illustrations through simulation are presented to demonstrate marginal improvements in using them rather than their predecessors. Two of the latter, though not optimal, are simpler, intuitively appealing, compete well in large samples, generally applicable and should be persisted with in practice.     

Unbiased statistical estimation functions for parameters in presence of nuisance parameters

We consider the problem of estimating a vector interesting parameter in the presence of nuisance parameters through vector unbiased statistical estimation functions (USEFs). An extension of the Cramer—Rao inequality relevant to the present problem is obtained. Three possible optimality criteria in the class of regular vector USEFs are those based on (i) the non-negative definiteness of the difference of dispersion matrices (ii) the trace of the dispersion matrix and (iii) the determinant of the dispersion matrix. We refer to these three criteria as M-optimality, T- optimality and D-optimality respectively. The equivalence of these three optimality criteria is established. By restricting the class of regular USEFs considered by Ferreira (1982), we study some interesting properties of the standardized USEFs and establish essential uniqueness of standardized M-optimal USEF in this restricted class. Finally some illustrative examples are included.     

On the optimality of a class of minimal covering designs

It is shown that the minimal covering designs for v=6t+5 treatments in blocks of size 3 are optimal w.r.t. a large class of optimality criteria. This class of optimality criteria includes the well-known criteria of A-, D- and E-optimality. It is conjectured that these designs are also optimal w.r.t. other criteria suggested by Takeuchi (1961).     

Compound designs for dose‐finding in the presence of nondesignable covariates

Compound optimal designs are considered where one component of the design criterion is a traditional optimality criterion such as the D‐optimality criterion, and the other component accounts for higher efficacy with low toxicity. With reference to the dose‐finding problem, we suggest the technique to choose weights for the two components that makes the optimization problem simpler than the traditional penalized design. We allow general bivariate responses for efficacy and toxicity. We then extend the procedure in the presence of nondesignable covariates such as age, sex, or other health conditions. A new breast cancer treatment is considered to illustrate the procedures. Copyright © 2013 John Wiley & Sons, Ltd.