Estimation of the Log Gaussian Cox Process Model with Point and Aggregated Case Data: The rts2 Package for R

Log Gaussian Cox Process (LGCP) models are a key method for disease and other types of surveillance, and provide a means of predicting risk across an area of interest based on spatially-referenced and time-stamped case data. However, these models can be difficult to specify and computationally demanding to estimate. For many surveillance scenarios we require results in near real-time using routinely available data to guide and direct policy responses, or due to limited availability of computational resources. There are limited software implementations available for this real-time context with reliable predictions and quantification of uncertainty. The package provides a range of modern Gaussian process approximations and model fitting methods to fit the LGCP, including estimation of covariance parameters, using both Bayesian and stochastic Maximum Likelihood methods. The package provides a suite of data manipulation tools. We also provide a new implementation to estimate the LGCP when case data are aggregated to an irregular grid such as census tract areas.