Premium
Scale problem in classification: an application of a stochastic method to evaluate the relative homogeneity of sample units
Author(s) -
DALE P. E. R.
Publication year - 1983
Publication title -
australian journal of ecology
Language(s) - English
Resource type - Journals
eISSN - 1442-9993
pISSN - 0307-692X
DOI - 10.1111/j.1442-9993.1983.tb01606.x
Subject(s) - homogeneity (statistics) , statistics , sample size determination , sample (material) , spatial ecology , efficiency , mathematics , ecology , scale (ratio) , econometrics , computer science , geography , biology , cartography , chemistry , chromatography , estimator
A basic problem in experimental design is to select a sample unit at a scale appropriate to the variables being observed and to the nature of the topic. Where hierarchical classification procedures are used it is important that the scale of the basic units is such that any variation found within the sample is small relative to that between the groups formed. In this paper a method is applied to evaluate the relative homogeneity of sample units. Replicates from each sample are classified hierarchically. The method of Sandland & Young (1979a, b) is used to determine the probability at division of the number of replicates assigned at a division to the same group occurring by chance. A high probability indicates relative heterogeneity of the samples from which replicates are drawn; a low probability suggests relative homogeneity. An example from the area of urban ecology illustrates an application of the method. It is used to evaluate the relative homogeneity of spatial units (Census Collectors Districts) with respect to dwelling characteristics in inner city Brisbane. Possible applications in other areas of ecology are briefly considered.