Geography Reference
In-Depth Information
0.07000
PP A
PP B
0.06000
0.05000
0.04000
0.03000
0.02000
0.01000
0.00000
0
5
10
15
20
25
30
35
Distance
Figure 7.11
KE for point patterns A and B for different bandwidths.
Clearly, the size of the kernel bandwidth will have a marked impact on the intensity
estimates. In practice, a variety of bandwidths is ot en used and the dif erence in results
assessed. In the cases described above, the size of the kernel bandwidth is the same at
all locations, but techniques exist such that the bandwidth can be varied from place to
place as a function of local event intensity, with a small bandwidth in areas with intense
point patterns and a large bandwidth in areas with a low event intensity (Bailey and
Gatrell, 1995; Brunsdon, 1995).
Measures based on distances between events
7.4
h e second-order properties of a point pattern (as dei ned in Section 7.1) can be mea-
sured using information on distances between events. h is section deals with such
approaches. h e initial concern is with methods that make use of information on the
nearest neighbours of events. h e second concern is with the
K
function, a widely used
means of assessing the degree of spatial dependence in point patterns (i.e. does the
number of points in one area tend to be similar to the number of points in another area
some i xed distance away?)
7.4.1
Nearest-neighbour methods
Various measures exist that are based on the nearest neighbour to each event. h e
mean nearest-neighbour distance is one such measure and it is given by:
=
Â
n
d
()
x
(7.9)
min
i
d
i
=
1
min
n
Search WWH ::
Custom Search