Model Selection Based On The Neyman-Scott Process

Spatial point processes started a little late in the theory of stochastic point processes,however,with the rapid development of computer technology and the development of GIS geographic information system,it developed quite rapidly,especially in the recent years.Its importance is increasingly prominent.Now Spatial point process has very frequent and successful application in the ecology,forestry,geography,spatial epidemiology and many other disciplines.The clustering point process is one of the most widely-used spatial distributions in the nature.And Neyman-Scott point process is one of the most popular probability models that can descript the spatial point distribution pattern.There are few studies on the testing and model selection of these stochastic processes.In this paper,we will share two methods of model selection for the Neyman-Scott point processes.The first method is to get the parameter estimation by minimizing the difference between the empirical K-function and the theoretical K-function of the model;and then choose the one with the smaller absolute error integral as the better model.The other method is to use RJMCMC algorithm based on bayesian factor theory to select the Neyman-Scott process model.and this is an improvement over the AIC criterion commonly used.In the end of this paper,the numerical simulation and example analysis of the two methods are carried out respectively,and the results are consistent with the existing results.