Almost everything we have discussed in this course has its analogue in a space-time setting. Methods for space-time data are less well developed than for purely spatial (or purely temporal) data, but this situation is changing rapidly
this course has included very little detailed discussion of inference for spatial data:
for exploratory analysis of case-control data, non-parametric methods are widely used, and hypothesis testing can be based on the randomisation distribution induced by the study design;
in other settings, parametric modelling assumptions are widely used, and inference uses likelihood-based methods, whether classical or Bayesian;
the Bayesian paradigm is particularly well suited to problems involving predictive inference for latent spatial processes, because it naturally adjusts for parameter uncertainty in the construction of prediction intervals.
software to implement most of the methods described in this course is freely available:
point patterns: spatstat, sp, splancs, soapp, spatialkernel
continuous spatial variation: gstat, automap, geoR, geoRglm.
discrete spatial variation: CARBayes, INLA, WinBUGS, BayesX. The last two are stand-alone programs; see
and
respectively.
The above R packages are available from the Comprehensive R Archive Network (CRAN): try searching for “spatstat CRAN” in google for instance.