Statistical Analysis Methods Page 3 |
Distance and Nearest Neighbours
A fundamental concept in the analysis of the point pattern data is distance. The (Euclidean) distance, d(i,j), between points i and j in 3D space with locations (or coordinates) (xi , yi , zi) and (xj , yj , zj) respectively is d(i,j) = √{( xi - xj )2 + ( yi - yj )2 + ( zi - zj )2} that is, the square-root of the squared differences between the x, y and z coordinates.
For a set of n data points corresponding to object locations, we have a total of N=n(n-1)/2 distances between objects taken pairwise. We can arrange the distances into a symmetric n x n matrix, D, with (i,j)th element d(i,j) and zeros on the diagonal.
The ith row of D contains the distances from object i to all other objects; the nearest neighbour of i is the object whose distance from i is smallest. That is, the nearest neighbour of i is denoted ηi and defined mathematically by
ηi = argminj d(i,j)
and the nearest neighbour distance is denoted δi and defined by
δi = d(i,ηi) = minj d(i,j)
that is, the smallest observed distance from object i.
Thus, for any data set of size n, we can extract a set of N inter-point distances, and a set of n nearest neighbour differences. These sets of distances can be used to test our hypotheses of interest.
|