24.07.2013 Views

Myriam Elizabeth Saavedra López - Repositorio Digital USFQ ...

Myriam Elizabeth Saavedra López - Repositorio Digital USFQ ...

Myriam Elizabeth Saavedra López - Repositorio Digital USFQ ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

for r ≥ 0 and in the general d-dimensional case, g (r) = K′ (r)<br />

dπdrd−1 for r ≥ 0.<br />

When r is large, the lim g (r) = 1, means there is independent between the two compared points<br />

r→∞<br />

for large r. To be more precise, for this equation, the point process must have some distributional<br />

properties (called mixing properties).<br />

If there is a finite distance rcorr with g (r) = 1 for r ≥ rcorr then rcorr is called range of correlation.<br />

This means that there are no correlations between point positions at larger distances.<br />

It is clear that in the Poisson process or CSR case g (r) = 1 for r ≥ 0, i.e. which means that the<br />

location of any point is entirely independent of the locations of the other points. Again, in typical<br />

non-Poisson cases a characteristic behaviour of g (r) may be found in the cluster process if g (r) > 1<br />

and regular process if g (r) < 1 , in particular for small radii in both cases.<br />

The Second Moment Characteristics are related to statistical concepts such as spatial autocorrelation<br />

and semi-variogram because all these functions study the distances (space) between variables (spatial<br />

data) and what is it happening between these variables when these distances are short or long (in-<br />

teraction and/or dependence). Therefore all these functions support the first law of geography that<br />

“everything is related to everything else, but near things are more related than distant things” (Waldo<br />

Tobler). See the following descriptions:<br />

1. Spatial Autocorrelation, according to Zuur, states that pairs of subjects (points) that are close<br />

to each other are more likely to have values that are more similar, and pair of subjects far apart<br />

forms each other are more likely to have values that are less similar. It is important to mention<br />

that the Function Ripley’s K is one of the indices of spatial autocorrelation.<br />

2. According to Cressei (1993, 58), semi-variogram has been called the variogram divided by 2,<br />

by Matheron (1962), therefore they are commonly referenced. The semi-variogram is a plot of<br />

semivariance as a function of distance. The semi-variance measures the dissimilarity of sub-<br />

jects within a single variable, compared to covariance which measures the similarity of one or<br />

more variables. It is not normalized and values are not as constrained as are most correlation<br />

coefficients (Zuur).<br />

3.6 Corrected Estimates<br />

According to Illian et al (2008, 180-183), the spatial statistics analysis with stationary point processes<br />

faces a difficult problem at its window edges: data are given for a bounded observation window W<br />

only, but the pattern is (implicitly) assumed to be infinite and the summary characteristic to be<br />

estimated is defined independently of W and should not show any traces of W . However, natural<br />

16

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!