Robust Gaussian stochastic process emulation. (English) Zbl 1408.62155

Gu, Mengyang; Wang, Xiaojing; Berger, James O.

Ann. Stat. 46, No. 6A, 3038-3066 (2018).

Summary: We consider estimation of the parameters of a Gaussian Stochastic Process (GaSP), in the context of emulation (approximation) of computer models for which the outcomes are real-valued scalars. The main focus is on estimation of the GaSP parameters through various generalized maximum likelihood methods, mostly involving finding posterior modes; this is because full Bayesian analysis in computer model emulation is typically prohibitively expensive.{ }The posterior modes that are studied arise from objective priors, such as the reference prior. These priors have been studied in the literature for the situation of an isotropic covariance function or under the assumption of separability in the design of inputs for model runs used in the GaSP construction. In this paper, we consider more general designs (e.g., a Latin Hypercube Design) with a class of commonly used anisotropic correlation functions, which can be written as a product of isotropic correlation functions, each having an unknown range parameter and a fixed roughness parameter. We discuss properties of the objective priors and marginal likelihoods for the parameters of the GaSP and establish the posterior propriety of the GaSP parameters, but our main focus is to demonstrate that certain parameterizations result in more robust estimation of the GaSP parameters than others, and that some parameterizations that are in common use should clearly be avoided. These results are applicable to many frequently used covariance functions, for example, power exponential, Matérn, rational quadratic and spherical covariance. We also generalize the results to the GaSP model with a nugget parameter. Both theoretical and numerical evidence is presented concerning the performance of the studied procedures.

MSC:

62M30	Inference from spatial processes
62F35	Robustness and adaptive procedures (parametric inference)
62K10	Statistical block designs

Keywords:

anisotropic covariance; emulation; Gaussian stochastic process; objective priors; posterior propriety; robust parameter estimation; Latin hypercube design

Software:

DiceOptim; DiceKriging; Virtual library

References:

[1]	An, J. and Owen, A. (2001). Quasi-regression. J. Complexity17 588–607. · Zbl 0993.65018 · doi:10.1006/jcom.2001.0588
[2]	Andrianakis, I. and Challenor, P. G. (2012). The effect of the nugget on Gaussian process emulators of computer models. Comput. Statist. Data Anal.56 4215–4228. · Zbl 1255.62306 · doi:10.1016/j.csda.2012.04.020
[3]	Bayarri, M., Berger, J., Cafeo, J., Garcia-Donato, G., Liu, F., Palomo, J., Parthasarathy, R., Paulo, R., Sacks, J. and Walsh, D. (2007). Computer model validation with functional output. Ann. Statist.35 1874–1906. · Zbl 1144.62368 · doi:10.1214/009053607000000163
[4]	Bayarri, M. J., Berger, J. O., Calder, E. S., Dalbey, K., Lunagomez, S., Patra, A. K., Pitman, E. B., Spiller, E. T. and Wolpert, R. L. (2009). Using statistical and computer models to quantify volcanic hazards. Technometrics51 402–413. · doi:10.1198/TECH.2009.08018
[5]	Berger, J. O., De Oliveira, V. and Sansó, B. (2001). Objective Bayesian analysis of spatially correlated data. J. Amer. Statist. Assoc.96 1361–1374. · Zbl 1051.62095 · doi:10.1198/016214501753382282
[6]	Dette, H. and Pepelyshev, A. (2010). Generalized latin hypercube design for computer experiments. Technometrics52 421–429.
[7]	De Oliveira, V. (2007). Objective Bayesian analysis of spatial data with measurement error. Canad. J. Statist.35 283–301. · Zbl 1129.62086 · doi:10.1002/cjs.5550350206
[8]	Diggle, P. and Ribeiro, P. (2007). Model-Based Geostatistics. Springer, Berlin. · Zbl 1132.86002
[9]	Dixon, L. (1978). The global optimization problem: An introduction. In Towards Global Optimiation2 1–15. North-Hollad, Amsterdam.
[10]	Gelfand, A. E. (2010). Handbook of Spatial Statistics. CRC Press, Boca Raton, FL. · Zbl 1188.62284
[11]	Gramacy, R. B. and Lee, H. K. (2009). Adaptive design and analysis of supercomputer experiments. Technometrics51 130–145.
[12]	Gu, M., Berger, J. O. et al. (2016). Parallel partial Gaussian process emulation for computer models with massive output. Ann. Appl. Stat.10 1317–1347. · Zbl 1391.62184 · doi:10.1214/16-AOAS934
[13]	Gu, M., Wang, X. and Berger, J. O. (2018). Supplement to “Robust Gaussian stochastic process emulation.” DOI:10.1214/17-AOS1648SUPP.
[14]	Handcock, M. S. and Stein, M. L. (1993). A Bayesian analysis of kriging. Technometrics35 403–410.
[15]	Handcock, M. S. and Wallis, J. R. (1994). An approach to statistical spatial-temporal modeling of meteorological fields. J. Amer. Statist. Assoc.89 368–378. · Zbl 0798.62109 · doi:10.1080/01621459.1994.10476754
[16]	Higdon, D. (2002). Space and space–time modeling using process convolutions. In Quantitative Methods for Current Environmental Issues 37–56. Springer, London. · Zbl 1255.86016
[17]	Kazianka, H. (2013). Objective Bayesian analysis of geometrically anisotropic spatial data. J. Agric. Biol. Environ. Stat.18 514–537. · Zbl 1303.62080 · doi:10.1007/s13253-013-0137-y
[18]	Kazianka, H. and Pilz, J. (2012). Objective Bayesian analysis of spatial data with uncertain nugget and range parameters. Canad. J. Statist.40 304–327. · Zbl 1348.62082 · doi:10.1002/cjs.11132
[19]	Kennedy, M. C. and O’Hagan, A. (2001). Bayesian calibration of computer models. J. R. Stat. Soc. Ser. B. Stat. Methodol.63 425–464. · Zbl 1007.62021 · doi:10.1111/1467-9868.00294
[20]	Li, R. and Sudjianto, A. (2005). Analysis of computer experiments using penalized likelihood in Gaussian kriging models. Technometrics47 111–120.
[21]	Linkletter, C., Bingham, D., Hengartner, N., Higdon, D. and Kenny, Q. Y. (2006). Variable selection for Gaussian process models in computer experiments. Technometrics48 478–490.
[22]	Morris, M. D., Mitchell, T. J. and Ylvisaker, D. (1993). Bayesian design and analysis of computer experiments: Use of derivatives in surface prediction. Technometrics35 243–255. · Zbl 0785.62025 · doi:10.1080/00401706.1993.10485320
[23]	Oakley, J. and O’Hagan, A. (2002). Bayesian inference for the uncertainty distribution of computer model outputs. Biometrika89 769–784.
[24]	Paciorek, C. J. and Schervish, M. J. (2006). Spatial modelling using a new class of nonstationary covariance functions. Environmetrics17 483–506.
[25]	Paulo, R. (2005). Default priors for Gaussian processes. Ann. Statist.33 556–582. · Zbl 1069.62030 · doi:10.1214/009053604000001264
[26]	Peng, C.-Y. and Wu, C. J. (2014). On the choice of nugget in kriging modeling for deterministic computer experiments. J. Comput. Graph. Statist.23 151–168.
[27]	Qian, P. Z. G., Wu, H. and Wu, C. F. J. (2008). Gaussian process models for computer experiments with qualitative and quantitative factors. Technometrics50 383–396.
[28]	Ranjan, H. R. and Karsten, R. (2011). A computationally stable approach to Gaussian process interpolation of deterministic computer simulation data. Technometrics53 366–378.
[29]	Ren, C., Sun, D. and He, C. (2012). Objective Bayesian analysis for a spatial model with nugget effects. J. Statist. Plann. Inference142 1933–1946. · Zbl 1237.62034 · doi:10.1016/j.jspi.2012.02.034
[30]	Ren, C., Sun, D. and Sahu, S. K. (2013). Objective Bayesian analysis of spatial models with separable correlation functions. Canad. J. Statist.41 488–507. · Zbl 1273.62032 · doi:10.1002/cjs.11186
[31]	Roustant, O., Ginsbourger, D. and Deville, Y. (2012). DiceKriging, DiceOptim: Two R packages for the analysis of computer experiments by kriging-based metamodeling and optimization. J. Stat. Softw.51 1–55.
[32]	Sacks, J., Welch, W. J., Mitchell, T. J. and Wynn, H. P. (1989). Design and analysis of computer experiments. Statist. Sci.4 409–435. · Zbl 0955.62619 · doi:10.1214/ss/1177012413
[33]	Santner, T. J., Williams, B. J. and Notz, W. I. (2003). The Design and Analysis of Computer Experiments. Springer, New York. · Zbl 1041.62068
[34]	Spiller, E. T., Bayarri, M., Berger, J. O., Calder, E. S., Patra, A. K., Pitman, E. B. and Wolpert, R. L. (2014). Automating emulator construction for geophysical hazard maps. SIAM/ASA J. Uncertain. Quantif.2 126–152. · Zbl 1308.62182 · doi:10.1137/120899285
[35]	Stein, M. L. (2012). Interpolation of Spatial Data: Some Theory for Kriging. Springer Science & Business Media, Berlin.
[36]	Surjanovic, S. and Bingham, D. (2017). Virtual library of simulation experiments: Test functions and datasets. Retrieved June 26, 2017, from http://www.sfu.ca/ ssurjano.
[37]	Zhang, H. (2004). Inconsistent estimation and asymptotically equal interpolations in model-based geostatistics. J. Amer. Statist. Assoc.99 250–261. · Zbl 1089.62538 · doi:10.1198/016214504000000241
[38]	Zhang, H. and Zimmerman, D. L. (2005). Towards reconciling two asymptotic frameworks in spatial statistics. Biometrika92 921–936. · Zbl 1151.62348 · doi:10.1093/biomet/92.4.921
[39]	Zimmerman, D. L. (1993). Another look at anisotropy in geostatistics. Math. Geol.25 453–470.