ANALYSIS OF RESIN COMPOSITIONAL DATA

Resin composition can be quantified with relative ease but, because of the properties of proportional data, analysis and interpretation of p atterns of variation in the resultant compositional data sets are rath er less easy. Proportions of a composition are constrained to sum to u nity; individual proportions are not, therefore, independent variables , and cannot be analyzed as such. Resin data shares with other composi tional data sets various analytical and interpretative difficulties: a major limitation for taxonomic, genetic and biosynthetic studies is t he absence of any interpretable covariance structure for proportional data sets. A valid approach to the interpretation of resin composition al data is through the analysis of ratios of proportions of a composit ion, which have the essential property of being invariant under reseal ing. Transformation of the proportional data to a logratio data set al lows the properties of the lognormal distribution to be employed; the logratio transformation removes the constraint of summation to unity, and provides an interpretable covariance structure. The logratios form a multivariate data set amenable to analysis using standard technique s, which we demonstrate for a sample set of resin compositional data.