Assessing minimum contrast parameter estimation for spatial and spatiotemporal log-Gaussian Cox processes

The univariate log‐Gaussian Cox process (LGCP) has shown considerable potential for the flexible modelling of the spatial, and more recently, spatiotemporal, intensity functions of planar point patterns within a restricted region in space. Its flexibility and mathematical tractability are partly offset by the need to acquire sensible estimates of the parameters controlling the dependence structure of the Gaussian field given the observed data. The method of minimum contrast, which compares theoretical descriptors of the process with their non‐parametric counterparts in order to obtain the required estimates, is arguably the most popular in practice to date. This article provides a comprehensive set of simulation studies focused on gauging the performance and impact of minimum contrast methods for parameter estimation of these processes. Results indicate that concerns over arbitrariness of implementation of minimum contrast give way to satisfactory practical performance.

[1]  Jiří Dvořák,et al.  Statistics for Inhomogeneous Space-Time Shot-Noise Cox Processes , 2014 .

[2]  R. Waagepetersen,et al.  Two‐step estimation for inhomogeneous spatial point processes , 2009 .

[3]  Peter J. Diggle,et al.  Corrigendum: Spatiotemporal prediction for log‐Gaussian Cox processes , 2003 .

[4]  P. Diggle,et al.  Spatiotemporal prediction for log‐Gaussian Cox processes , 2001 .

[5]  John A. Nelder,et al.  A Simplex Method for Function Minimization , 1965, Comput. J..

[6]  Peter J. Diggle,et al.  lgcp: An R Package for Inference with Spatial and Spatio-Temporal Log-Gaussian Cox Processes , 2013 .

[7]  Peter J. Diggle,et al.  Erratum: Spatiotemporal prediction for log-Gaussian Cox processes (Journal of the Royal Statistical Society. Series B: Statistical Methodology (2001) 63 (823-841)) , 2003 .

[8]  Jesper Møller,et al.  Structured Spatio‐Temporal Shot‐Noise Cox Point Process Models, with a View to Modelling Forest Fires , 2010 .

[9]  Peter J. Diggle,et al.  Lgcp: Inference with spatial and spatio-temporal log-gaussian cox processes in R , 2013 .

[10]  Yongtao Guan.  A minimum contrast estimation procedure for estimating the second-order parameters of inhomogeneous spatial point processes , 2009 .

[11]  A. Baddeley,et al.  Non‐ and semi‐parametric estimation of interaction in inhomogeneous point patterns , 2000 .

[12]  R. Waagepetersen,et al.  Bayesian Prediction of Spatial Count Data Using Generalized Linear Mixed Models , 2002, Biometrics.

[13]  B. Ripley.  Modelling Spatial Patterns , 1977 .

[14]  Yongtao Guan.  A least-squares cross-validation bandwidth selection approach in pair correlation function estimations , 2007 .

[15]  P. Diggle,et al.  Monte Carlo Methods of Inference for Implicit Statistical Models , 1984 .

[16]  J. Møller,et al.  Shot noise Cox processes , 2003, Advances in Applied Probability.

[17]  Peter J. Diggle,et al.  Point process methodology for on‐line spatio‐temporal disease surveillance , 2005 .

[18]  David Bryant,et al.  On Circulant Embedding for Gaussian Random Fields in R , 2013 .

[19]  P. Diggle,et al.  Spatial variation in risk of disease: a nonparametric binary regression approach , 2002 .

[20]  Adrian Baddeley,et al.  spatstat: An R Package for Analyzing Spatial Point Patterns , 2005 .

[21]  Yongtao Guan,et al.  On least squares fitting for stationary spatial point processes , 2007 .

[22]  P. Lewy,et al.  Modelling the distribution of fish accounting for spatial correlation and overdispersion , 2009 .

[23]  Kim Mouridsen,et al.  Lévy‐based Modelling in Brain Imaging , 2013 .

[24]  Anthony N. Pettitt,et al.  Binary probability maps using a hidden conditional autoregressive Gaussian process with an application to Finnish common toad data , 2000 .

[25]  R. Tweedie,et al.  Exponential convergence of Langevin distributions and their discrete approximations , 1996 .

[26]  J. Møller,et al.  Log Gaussian Cox Processes , 1998 .