×

Investigating group-specific clustering in univariate spatial data with many small groups. (English) Zbl 0558.62092

In a situation where no assumption of stationarity or distributional form can be made, the problem is to investigate whether there tend to be local clusters within groups of observations. The null hypothesis is that the points are grouped randomly. One method is to compare the distribution of interpoint distances within groups with that in the pooled sample. Another method is to look at the rank distances between neighbouring points within groups, the ranks being obtained from the pooled sample. Many small rank distances indicate group-specific clustering. These methods are applied to a medical problem.

MSC:

62P10 Applications of statistics to biology and medical sciences; meta analysis
62G10 Nonparametric hypothesis testing
PDFBibTeX XMLCite
Full Text: DOI