Title: New quantitative description for spatial point patterns
An important task of the analysis
of spatial point patterns is to examine the dependence between points.
Many summary functions, for example, the nearest neighbour distance
distribution, Ripley's K-function, the empty space function and
the periodogram, have been proposed to extract such information.
One way to assess how good a summary function can capture the dependence
structure of a given point pattern is to see how sensitive it will be when
it is used to test the complete spatial randomness hypothesis against various
alternatives. A point pattern is said to be completely spatially
random if it is a realisation of a stationary Poisson process or a binomial
process. Comparative studies, however, suggest that there is no uniformly
best test. In this talk we will introduce new quantitative measures
that are more able to detect certain kinds of dependence structure than
the existing summary functions.