This course introduces statistical approaches to modelling data that have a spatial structure. The course will focus on modelling approaches for the three main types of spatially orientated data, namely: (i) geostatistics; (ii) areal (lattice) data; and (iii) spatial point processes.

Lectures: 20

Tutorials: 5

Practical: 2, two-hour computer lab sessions

STATS4xxx (corresponding H level course)

120-minute, end-of-course examination (100%)

To introduce the three main types of spatial data and describe how:

to identify trends and spatial autocorrelation.

to model spatial autocorrelation.

to predict the spatial process at unmeasured locations.

to apply the methodology to real spatial data sets in the statistical package R.

to use spatial modelling techniques in a research context.

By the end of this course students will be able to:

■ describe the differences between geostatistical data, areal unit data and point process data.

■ use descriptive techniques to determine whether spatially structured data exhibit spatial autocorrelation.

■ define the concepts of stationarity and isotropy.

■ define the nugget, range and sill parameters of a spatial autocorrelation functions.

■ derive the Kriging equations for spatial prediction.

■ define and derive the class of conditionally autoregressive models.

■ determine whether a spatial point process has any spatial structure.

■ use the statistical package R to fit appropriate spatial models to geostatistical data, areal unit data and point process data.

■ Critique research papers in spatial statistics.

None