Mackay and Oldford’s perspective is perhaps a little narrow and somewhat less exploratory and iterative than applies for many complex real-world problems. In respect of problems that the GIS user is likely to meet there are a number of reasons for these differences, most notably:

·         Spatial analysis is particularly concerned with problems that have an explicit spatial context, and frequently data at one location is not independent of data at other locations (see Sections 2.3 and 2.4). Indeed such associations (spatial correlations) are the norm, especially for measurements taken at locations that are near to one another. Identifying and analysing such patterns is often the goal of analysis, at least at the early stages of investigation

·         Many problems must be considered in a spatio-temporal context rather than simply a spatial context. Time of day/week/month/year may have great relevance to obtaining an understanding of particular spatial problems, especially those of an environmental nature and problems relating to infrastructure usage and planning

·         The theoretical foundations of statistics rely on a set of assumptions and sampling procedures that are often more applicable to experimental datasets than purely observational data. Very few problems addressed by spatial analysis fall into the category of truly experimental research

·         Often the purpose of spatial analysis is not merely to identify pattern, but to construct models, if possible by gaining an understanding of process. But spatial patterns are rarely if ever uniquely determined by a single process, hence spatial analysis is often the start of further investigations into process and model building, and rarely an end in itself (see also, the discuss on Agent-based modelling in Section 8.2.7). Where explicitly spatial factors, such as distance or contiguity, are identified as important or significant, the question why? must follow: is the problem under consideration directly or indirectly affected by purely spatial factors; or is the spatial element a surrogate for one or more explanatory variables that have not be adequately modelled or are unobtainable?

·         Spatial datasets are often provided by third parties, such as national mapping agencies, census units and third party data vendors. Metadata provided with such material may or may not provide adequate information on the quality, accuracy, consistency, completeness and provenance of the information. In many areas of spatial research these elements are pre-determined, although they are often augmented by corporate datasets (e.g. customer databases, crime incident records, medical case details) or field research (e.g. georeferenced collection of soil samples or plant locations, market research exercises, bathymetric surveys etc.)

Each of these factors serves to distinguish spatial analysis from analysis in other disciplines, whilst at the same time recognising the considerable similarities and overlap of methodologies and techniques. In subsections 3.2.3.1 to 3.2.3.5 we examine each step of the PPDAC model in the context of spatial analysis. In this “revised” version of the PPDAC approach the Plan stage is much broader than in Mackay and Oldford’s model. In their approach the Plan stage focuses largely on the data collection procedures to be adopted. In our case it covers all aspects of preparation for data acquisition and analysis, including considerations of feasibility within given project constraints.

For many problems that arise in spatial analysis there will be multiple instances of the process described, especially for the data and analysis stages, where several different but related datasets are to be studied in order to address the problem at hand. Finally, in instances where the geospatial analyst is simply presented with the data and asked to carry out appropriate analyses, it is essential that the context is first understood (i.e. Problem, Plan and Data) even if this cannot be influenced or revisited.