This work focuses on combining observations from field experiments with detailed computer simulations of a physical process to carry out statistical inference. Of particular interest here is determining uncertainty in resulting predictions. This typically involves calibration of parameters in the computer simulator as well as accounting for inadequate physics in the simulator. The problem is complicated by the fact that simulation code is sufficiently demanding that only a limited number of simulations can be carried out. We consider applications in characterizing material properties for which the field data and the simulator output are highly multivariate. For example, the experimental data and simulation output may be an image or may describe the shape of a physical object. We make use of the basic framework of Kennedy and O'Hagan (2001). However, the size and multivariate nature of the data lead to computational challenges in implementing the framework. To overcome these challenges, we make use of basis representations (e.g. principal components) to reduce the dimensionality of the problem and speed up the computations required for exploring the posterior distribution. This methodology is applied to applications, both ongoing and historical, at Los Alamos National Laboratory.
We develop a statistical approach for characterizing uncertainty in predictions that are made with the aid of a computer simulation model. Typically, the computer simulation code models a physical system and requires a set of inputs-some known and specified, others unknown. A limited amount of field data from the true physical system is available to inform us about the unknown inputs and also to inform us about the uncertainty that is associated with a simulationbased prediction. The approach given here allows for the following: • uncertainty regarding model inputs (i.e., calibration); • accounting for uncertainty due to limitations on the number of simulations that can be carried out; • discrepancy between the simulation code and the actual physical system; • uncertainty in the observation process that yields the actual field data on the true physical system. The resulting analysis yields predictions and their associated uncertainties while accounting for multiple sources of uncertainty. We use a Bayesian formulation and rely on Gaussian process models to model unknown functions of the model inputs. The estimation is carried out using a Markov chain Monte Carlo method. This methodology is applied to two examples: a charged particle accelerator and a spot welding process.
One of the main aims in epidemiology is the estimation and mapping of spatial variation in disease risk while adjusting for available covariate information. We analyse the spatial distribution of infant mortality cases compared to live-born controls from Porto Alegre, Rio Grande do Sul in a binary spatial regression model. A commonly used approach for such data is a spatial point process. Here the risk measure, which continously varies over the study region is estimated using generalized additive models, as presented by Kelsall and Diggle (1998). In this models, a smoothing parameter has to be chosen, which strongly affects the estimation of the spatial risk surface, and inference for the regression coefficients is conditional on the surface. Instead we discretise space on a square grid and use higher order Gaussian Markov random fields (GMRF). Some of the GMRF models are already described in Besag and Higdon (1999), we expanded their approach up to a GMRF for twelve nearest neighbours. In contrast to the spatial point process approach, the smoothing parameter in this model can be estimated and covariate effects and the spatial surface can be estimated jointly. However the level of discretisation, the choice of the GMRF and the hyper prioris affect the estimated surface, as we show. We follow a Bayesian binary regression approach using a data augmentation technique described by Holmes and Held (2003). Here the resulting conditional likelihood of the regression coefficients and the spatial component is multivariate normal. This allows for efficient simulation using block Gibbs sampler as in Knorr-Held and Rue (2001). The results shows duration of gestation and weight of the new born child as only significant covariates. The spatial analysis shows two regions with significantly higher mortality risk, not explained by the covariates. The results agree well with those presented in Shumakura et.al., who analyse the same data using a spatial point process approach. The Hamburg Leukaemia and Lymphoma study has the aim to investigate the influence of selected environmental emission sources on the incidence of leukaemia, lymphoma and plasmocytoma. As a part of this study the hypothesis of a possible contribution of traffic immissions was analysed. Period of observation was 1988-1999; the case data set (Cases = 10034) was collected by the Hamburg Cancer Registry (additional program for completion). As the control group a 10% sample of Hamburg population was drawn from the residence registry (Controls = 173177). The residential adress was coded to geographic coordinates (precision ± 5 [m]). Indication for exposure to traffic related immission was calculated by (a) a local immission prognosis as the sum of traffic induced hydrocarbons (HC µg/m³, grids 500*500 meter) and (b) by the number of main streets/highways within 250 m distance to the residential adress. Age, sex and exposure stratified calculation as well as unconditional logistic regression methods were applied to analyse the influence structure. For all malignant h...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.