Introduction
The project intends to use traditional
statistics (regression and sampling) to tackle the relationship between
the change in low-density residential land use and the change in individual
households' tax burden. However, dealing with the spatial aspects of the
project helps understand the complexity of the data in order to increase
the significance of the result of future regression models.
In fact, the special nature of spatial data, due
to the fact that the location of the observations provides important information,
requires appropriate methodologies for their analysis, developed in the
fields of spatial statistics and spatial econometrics. The basic difference
between traditional statistics and spatial statistics is the traditional
assumption of independent units of analysis, which in fact are spatially
autocorrelated in some ways in the case of the project where MCDs are units
of analysis.
ESDA (Exploratory spatial data analysis)
is concerned with the detection, analysis, and interpretation of spatial
patterns in the data, such as spatial clusters, outliers, and hot spots.
Several measures of spatial autocorrelation can be treated as well. Spatial
regression analysis deals with the effects of the special nature of geographic
data on the properties of regression models. One aspect of this is the
detection of spatial autocorrelation as a specification error in regression
models.
At the current stage of the project, bi-variate
thematic mapping and spatial autocorrelation mapping is used to observe
if there is any spatial pattern within the dataset. Mapping as well as
animation are used to observe spatial intervening factors that can not
be taken into account in traditional regression models such as the effect
of highways and sewer zones on the location of new land uses and the location
of a MCD within a regional context. This will facilitate the ESDA and spatial
regression analysis later on in SpaceStat.
|