Bayesian inference for the fitting of dry matter accumulation curves in garlic plants

Abstract

The objective of this work was to identify nonlinear regression models that best describe dry matter accumulation curves over time, in garlic (Allium sativum) accessions, using Bayesian and frequentist approaches. Multivariate cluster analyses were made to group similar accessions according to the estimates of the parameters with biological interpretation (β1 and β3). In order to verify if the obtained groups were equal, statistical tests were applied to assess the parameter equality of the representative curves of each group. Thirty garlic accessions were used, which are kept by the vegetable germplasm bank of Universidade Federal de Viçosa, Brazil. The logistic model was the one that fit best to data in both approaches. Parameter estimates of this model were subjected to the cluster analysis using Ward’s algorithm, and the generalized Mahalanobis distance was used as a measure of dissimilarity. The optimal number of groups, according to the Mojena method, was three and four, for the frequentist and Bayesian approaches, respectively. Hypothesis tests for the parameter equality from estimated curves, for each identified group, indicated that both approaches highlight the differences between the accessions identified in the cluster analysis. Therefore, both approaches are recommended for this kind of study.