A New Class of IMSE-Based Criteria for Optimal Designs in Multi-response Random Coefficient Regression Models

Abstract

A new class of criteria for optimal designs in random coefficient regression (RCR) models with r responses is presented, which is based on the integrated mean squared error (IMSE) for the prediction of random effects. This class, referred to as

{\text{IMSE}}_{r,L}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\textrm{IMSE}_{r,L}$$\end{document}

-class of criteria, is invariant with respect to different parameterizations of the model and contains

\text{IMSE}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\textrm{IMSE}$$\end{document}

- and G-optimality as special cases for the prediction in univariate response situations. General equivalence theorems for

{\text{IMSE}}_{r,L}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\textrm{IMSE}_{r,L}$$\end{document}

-criteria are established for

L\in [1,\infty )

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$L\in [1,\infty )$$\end{document}

and

L=\infty

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$L=\infty $$\end{document}

, respectively, which are used to check

{\text{IMSE}}_{r,L}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\textrm{IMSE}_{r,L}$$\end{document}

-optimality of designs.

{\text{IMSE}}_{r,L}

\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\textrm{IMSE}_{r,L}$$\end{document}

-optimal designs for linear and quadratic bi-response RCR models are given for illustration.

Keywords

Optimal designs / Integrated mean squared error matrix / IMSE-optimality / Prediction / Mixed effects model

References

[1] ChristensenRPlane Answers to Complex Questions: The Theory of Linear Models, 2002New YorkSpringer.

[2] DanskinJMThe Theory of Max–Min and Its Application to Weapons Allocation Problems, 1967New YorkSpringer.

[3] DetteH, O’BrienTE. Optimality criteria for regression models based on predicted variance. Biometrika, 1999, 86: 93-106.

[4] FedorovVV, HacklPModel-Oriented Design of Experiments, 1997New YorkSpringer.

[5] GladitzJ, PilzJ. Construction of optimal designs in random coefficient regression models. Statistics, 1982, 13: 371-385

[6] HendersonCR, KempthorneO, SearleSR, von KrosigkCM. The estimation of environmental and genetic trends from records subject to culling. Biometrics, 1959, 15: 192-218.

[7] HendersonCR. Best linear unbiased estimation prediction under a selection model. Biometrics, 1975, 31: 423-477.

[8] JensenKL, SpiildH, ToftumJ. Implementation of multivariate linear mixed-effects models in the analysis of indoor climate performance experiments. Int. J. Biometeorol., 2012, 56: 129-136.

[9] KrafftO, SchaeferM. D\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$D$$\end{document}-optimal designs for a multivariate regression model. J. Multivar. Anal., 1992, 42: 130-140.

[10] LairdNM, WareJH. Random-effects models for longitudinal data. Biometrics, 1982, 38: 963-974.

[11] LiuX, YueR-X, HickernellFJ. Optimality criteria for multiresponse linear models based on predictive ellipsoids. Stat. Sin., 2011, 21: 421-432

[12] LiuX, YueR-X, LinDK. Optimal design for prediction in multiresponse linear models based on rectangular confidence region. J. Stat. Plan. Inference, 2013, 143: 1954-1967.

[13] LiuX, YueR-X, WongWK. D\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$D$$\end{document}-optimal designs for multi-response linear mixed models. Metrika, 2019, 82: 87-98.

[14] LiuX, YueRX, ChatterjeeK. Geometric characterization of D\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$D$$\end{document}-optimal designs for random coefficient regression models Statist. Prob. Lett., 2020, 159. 108696

[15] MasoudiE, HollingH, WongWK. Application of imperialist competitive algorithm to find minimax and standardized maximin optimal designs. Comput. Stat. Data Anal., 2017, 113: 330-345.

[16] PilzJBayesian Estimation and Experimental Design in Linear Regression Models, 1983LeipzigTeubner

[17] PinheiroJC, BatesDMMixed Effects Models in S and S-Plus, 2000New YorkSpringer.

[18] Prus, M.: Optimal designs for the prediction in hierarchical random coefficient regression models. Ph.D. Thesis. Otto-von-Guericke University, Magdeburg (2015)

[19] PrusM, SchwabeR. Optimal designs for the prediction of individual parameters in hierarchical models. J. R. Stat. Soc. B, 2016, 78: 175-191.

[20] PrusM. Various optimality criteria for the prediction of individual response curves. Stat. Probab. Lett., 2019, 146: 36-41.

[21] PrusM. Optimal designs for minimax-criteria in random coefficient regression models. Stat. Papers, 2019, 60: 465-478.

[22] PukelsheimFOptimal Design of Experiments, 1993New YorkWiley

[23] SchmelterT. The optimality of single-group designs for certain mixed models. Metrika, 2007, 65: 183-193.

[24] SpjøtvollE. Random coefficients regression models: a review. Math. Operforsch. Stat. Ser. Stat., 1977, 8: 69-93

[25] VerbekeG, MolenberghsGLinear Mixed Models for Longitudinal Data, 2000New YorkSpringer

[26] WangCM, LamCT. A mixed-effects model for the analysis of circular measurements. Technometrics, 1997, 39: 119-126.

[27] WhittleP. Some general points in the theory of optimal experimental designs. J. R. Stat. Soc. Ser. B, 1973, 35: 123-130.

[28] WongWK. A unified approach to the construction of minimax designs. Biometrika, 1992, 79: 611-619.

[29] WongWK, CookRD. Heteroscedastic G\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$G$$\end{document}-optimal designs. J. R. Stat. Soc. Ser. B, 1993, 55: 871-880.

Funding

National Natural Science Foundation of China(11971318)

Natural Science Foundation of Anhui Province (CN)(2008085QA15)

Shanghai Rising-Star Program(20QA1407500)

RIGHTS & PERMISSIONS

School of Mathematical Sciences, University of Science and Technology of China and Springer-Verlag GmbH Germany, part of Springer Nature