A note on Burnaby’s character-weighted similarity coefficient

Some of the problems of weighting characters and of eliminating correlational effects between characters are discussed. Many different weighting schemes can be proposed for similarity coefficients even when, as in Burnaby’s coefficient, it is decided to give high weight to matches between rare states. The rarity of a character state must be considered relative to the sampling scheme used to select the sampling units. Similarities between all pairs of units may be analyzed to suggest groups among the units. Groups having many correlated characters may be the most useful but are unlikely to be found by numerical methods which eliminate correlational effects.