Multivariate selection and classification of mathematical models to estimate dry matter partitioning in the fertigated Prata banana in the Northeast Brazil

Abstract

Abstract The selection of models to estimate dry matter partitioning for any crop is an important tool, since it enables rationalisation in the use of inputs and improvement in the management of an area, with gains in productivity. The aim was to use and compare different statistical evaluators and cluster analysis in the classification and selection of models with goodness of fit to estimate dry matter partitioning in banana mat. In the field, sixteen banana mats were collected ready for harvest and separated into mother plant (rhizome, pseudostem, leaf, stalk and fruit) and daughter plant (daughter rhizome, daughter pseudostem and daughter leaf). Models for the estimation of dry matter weight were generated for the different plant organs. Mean square residual (MSR), standard error of the estimate (Syx%), residual mean absolute deviation (MAD), coefficient of determination (R2), corrected Akaike information criterion (AICc) and Bayesian information criterion (BIC), in addition to cluster analysis were used for comparison and selection of the most appropriate model. The dry matter weight (DMw) of the organs comprising the mother plant and the daughter plant are best estimated as a function of DMw_Mother and DMw_Daughter respectively, except for the stalks and fruit. MSR, Syx%, MAD, R2, AICc and BIC are efficient tools in the selection and classification of models to estimate DMw partitioning in banana. Cluster analysis confirms the models selected by the evaluators, but unlike these, it selects only one as the “best model” for each plant organ.