The premises underlying univariate (CVA = critical value approach) and bivariate (DRIS = diagnosis and recommendation integrated system) diagnostic systems were reexamined with regard to compositional data analysis (CDA). CDA recognizes a structure of dependence among plant nutrients, the bounded sum constraint to one (the whole composition equals 100% or 1), and removes the curvature problem carried by crude components and by dual ratios or logratios when treated in isolation. Linearization by “rowcentered logrationing” of nutrient fractions shows great potential for carrying multivariate diagnosis and principal component analysis on nutrient data. Compositional nutrient diagnosis (CND) is supported by the theory of CDA. CND is the multivariate expansion of CVA and DRIS and is fully compatible with PCA. CND takes all possible nutrient interactions into account. CND nutrient indices are composed of two separate functions, one considering differences between nutrient levels, another examining differences between nutrient balances (as defined by nutrient geometric means), of individual and target specimens. These functions indicate that nutrient insufficiency can be corrected by either adding a single nutrient or taking advantage of multiple nutrient interactions to improve nutrient balance as a whole. A theoretical interpretative table is presented for CND.
If the inline PDF is not rendering correctly, you can download the PDF file here.