Using Mixtures-of-Distributions models to inform farm size selection decisions in representative farm modelling

Abstract

The selection of �representative� farms in farm level modelling where results are aggregated to the sector level is critically important if the effects of aggregation bias are to be reduced. The process of selecting representative farms normally involves the use of cluster analysis where the decision regarding the appropriate number of clusters (or representative farm types) is largely subjective. However, when the technique of fitting mixtures of distributions is employed as a clustering technique there is an objective test of the appropriate number of clusters. This paper demonstrates the MDM approach to cluster analysis by classifying dairy farms in Northern Ireland, based on the number of cows in each farm. The results indicate that four representative farms are needed, with a view to minimising aggregation bias, to describe the dairy sector in Northern Ireland.