Summary - Spatial Data Analysis: An Introduction for GIS Users - page 173

selection of appropriate methods is ot en not straightforward. Since there are many
alternative approaches for the spatial analyst to choose from, ot en experience is
important. As an example, there are numerous methods for spatial interpolation (the
focus of Chapter 9) and choice of specii c method, interpolation neighbourhood (use
all data or a local subset), and use of exact or approximate interpolation, are all likely
to be important. Guidance has been of ered in the text, although this can be no substi-
tute for experimentation and direct assessment of the dif erences in results obtained
using dif erent combinations of approaches. Descriptions of other methods are provided
in the publications referenced in Section 11.4.
Expensive technology and sophisticated-looking sot ware cannot disguise the fact
that spatial data analysis is based on models that may be poor abstractions of reality.
h ese limitations and the various kinds of errors that af ect any data source must be
considered. h e propagation of errors from one stage of processing to another (as referred
to in Section 2.9.1) is the subject of much research (see Burrough and McDonnell,
1998) and all users of (spatial) data are obliged to consider the quality of their data and
possible impacts on analyses that are based on these data.

Where next?

11.4

GIS and spatial data analysis are practical topics. If the tools of ered are not used then
there is no point in them. To begin to develop an understanding of how and why
particular approaches are employed and to become aware of their shortcomings, it is
necessary to apply the methods. In short, experimentation is a vital part of the learn-
ing process. h e topic website provides details of some sot ware packages that imple-
ment the methods described in the topic. Commercial packages like ArcGIS™ include
all of the basic analytical functions that most users are likely to need. Indeed almost all
of the case studies in this topic can be directly replicated using ArcGIS™ and its associ-
ated extensions. Some very extensive and powerful packages are completely free and
there is, therefore, not necessarily any i nancial barrier to sophisticated spatial data anal-
ysis. For example, the GIS GRASS (Neteler and Mitásová, 2007) and the R program-
ming environment 1 (Bivand et al ., 2008) and associated routines of er sophisticated
functionality at no cost.
In terms of reading material, there are several topic chapters and full topics that may
make sensible next steps. Brief summaries of methods for the analysis of spatial data which
provide pointers to more extensive material are provided by Anselin (2005), Getis (2005),
Fischer (2005), and Charlton (2008). h e topic by O'Sullivan and Unwin (2002) expands
on some of the issues discussed here. Bailey and Gatrell (1995), Lloyd (2006), and
De Smith et al. (2007) provide further accounts of methods discussed in this topic.
Many topics have been written for particular audiences. A good example is the topic by
Plane and Rogerson (1994), which deals with the analysis of data about human

1 http://www.r-project.org/