A Box–Cox transformed Gaussian process model for stochastic computer experiments
Ting Cui, Shifeng Xiong · Communications in Statistics - Simulation and Computation · 2025
In this paper we propose a Box–Cox transformed Gaussian process model for stochastic computer experiments. With this transformation, the proposed model can flexibly accommodate the non-normality of random errors. A plug-in approach and a Bayesian approach are presented to give prediction results under the model. To reduce the computational burden for large-scale datasets, we apply three computation-reduction methods, the Nyström method, the random Fourier features method, and the reconstruction parameterization method, to our approaches, and compare their prediction performance via Monte Carlo simulations. A real metro simulation dataset is also analyzed using the proposed methods.