A Typology of Polish Farms Using Probabilistic D-Clustering

The Agricultural Census conducted in Poland in 2010 was partially based on administrative sources. These data collection will be supplemented by sample survey of agricultural farm. This research is aimed at creation of an effective typology of Polish farms, which is necessary for proper sampling and reflection of many special types of agricultural activity, such as combining it with non. agricultural work. We propose some universal form of such typology constructed using data collected from administrative sources during the preliminary agricultural census conducted in autumn 2009. It is based on the especially prepared method of fuzzy clustering, i.e. probabilistic d-clustering adopted for interval data. For this reason, and because of an ambiguous impact of some key variables on classification, relevant criterions are presented as intervals. They are arbitrarily established, but also - as an alternative way - are generated endogenically, using an original optimization algorithm. For a better comparison, relevant classification for data collected “from nature” is provided.