Partition of information

Abstract

The classification of quantitative ecological survey data has in the past encountered two serious difficulties: (i) no completely satisfactory information statistic model has been devised for this purpose, and the theoretically most attractive - the "diversity" model - appears usually to produce ecologically meaningless fusions; (ii) no satisfactory partition of information between qualitative and quantitative components has been known. The requirement is for an information statistic model which can be confidently used for either qualitative or quantitative data, and which can be partitioned into its qualitative and quantitative components. A new statistic is defined which meets these requirements, and its application to a small-scale test example is illustrated.