Fundamental properties are inherent to the nature
of attributes as they are distributed across the earth’s surface. There is a
fundamental continuity (structure) to attributes in space that derives from the
underlying processes that shape the human and physical geographical world.
Continuity is also a fundamental property of attributes observed in time. If we
know the level of an attribute at one position in space (time) we can make an
informed estimate of its level at adjacent locations (points in time). Spatial
autocorrelation, in statistical terms, is a second order property of an
attribute distributed in geographic space. In addition there may be a mean or
first-order component of variation represented by a linear, quadratic, cubic
(etc.) trend. We can think of these as two different scales of spatial
variation although the distinction may be hard to make and quantify in
practice. As Cressie (1991) remarks: ‘What is one person’s (spatial) covariance
may be another persons mean structure’ (p. 25). It has often been remarked that
spatial variation is heterogeneous. This type of decomposition (plus a white
noise element to capture highly localized heterogeneity) is one way of formally
capturing that heterogeneity using what are termed ‘global’ models. Another
approach is to only analyze spatial subsets, that is allow model structure to
vary locally.
(Haining, R. 2009. The Special Nature of Spatial
Data (Chapter 2). Spatial Analysis (Handbook). Ed. A.S. Foteringham and P.A. Rogerson.
Sage Publications. 6 p.)
Types of GIS Spatial Data
In GIS, spatial data is classified as three main types: point, line,
and polygon.
A point is a convenient visual symbol (an X,
dot or other graphic), but it does not reflect the real dimensions of the
feature. Points may indicate specific locations (such as a given address, or
the occurrence of an event) and/or which are usually too small to depict properly
at the chosen scale features (such as a building).
A line is a one-dimensional feature with a
starting and an ending point. Lines represent linear features, either real (e.g., roads
or streams) as in Figure 2.2, or fictitious (e.g., administrative boundaries).
A polygon is an enclosed area, a
two-dimensional feature with at least three sides (and therefore with an area). For example, it
may represent a parcel of land, agricultural fields, or a political district.
(Fundamentals of GIS Data, Chapter 2, p.1
accessed at http://igre.emich.edu/wsatraining/TManual/Chapter2/Chap2.pdf)
What makes the analysis of spatial data
special is the fact that it has always played a central role in the
quantitative scientific tradition in geography. In general terms, spatial
analysis can be considered to be the formal quantitative study of phenomena
that manifest themselves in spare. This implies a focus on location, area,
distance and interaction, e.g., as expressed in Tobler's (1979) First Law of
Geography, where "everything is related to everything else, but near
things are more related than distant things." In order to interpret what
"near" and "distant" mean in a particular context,
observations on the phenomenon of interest need to be referenced in space,
e.g., in terms of points, lines or areal units.
The wide array of
philosophical and methodological dilemmas that confront the analysis of spatial
data necessitates an eclectic perspective. Many different ways of looking at a
data set or at a model specification should be compared, and sensitivity
analysis should play a central role. If different approaches yield the same
conclusions, one can be fairly confident that meaningful insights have been
gained. On the other hand, if the statistical findings turn out to be very
sensitive to the approach taken, there is likely to be something wrong with the
data and/or with the model and not much faith should be put in the precise
quantitative results.
The characteristics of errors that affect
observations of spatial data clearly motivate the need for a specialized
methodology of spatial statistics and spatial econometrics.
(Anselin, L. 1989. What’s Special about Spatial Data? Spring 1989
Symposium on Spatial Statistics,
Past, Present and Future,
Department of Geography, Syracuse University.)