PAIRWISE PARTITIONING - A NONMETRIC ALGORITHM FOR IDENTIFYING FEATURE-BASED SIMILARITY STRUCTURES

Pairwise partitioning is a nonmetric, divisive algorithm, for identify ing feature structures based on pairwise similarities. For errorless d ata, this algorithm is shown to identify only (and sometimes all) vali d features for certain hierarchical and multidimensional feature struc tures. Unfortunately, the algorithm is also extremely sensitive to err or in the data. Fortunately, several modifications of the algorithm ar e shown to compensate significantly for this deficiency. The algorithm is illustrated with simulations and analyses of three sets of similar ity data. The results suggest that this algorithm will be most useful (a) as an exploratory tool for generating a relatively large set of po tential features that can be reduced using other criteria (either stat istical or substantive) and (b) as a source of confirmatory or disconf irmatory evidence.