ABSTRACT
This model enabled also to infer that the clones FDR 5788, CDC312 and CDC56 presented high resitance, while the colne FX 3864 was the most susceptible in all studied locals. South American leaf blight caused by the fungus Microcyclus ulei is the most serious disease of rubber plantations in Latin America. With the aim of identifying more resistant rubber-tree clones in various environments over time, different random regression models (MRA) adjusted via a Bayesian approach were compared. Eight clones were tested in clonal fields in a randomised complete block design with four replications, using 80 trees per lot. The two central rows were evaluated every two months for the variables severity (SEV) and stroma index of the adult leaf (EFA). Trunk circumference and the climatic variables of each clonal field were included in the models as covariates. The MRA were compared by the deviance information criterion (DIC). The M2 model (which assumed quadratic random effects for clones and linear effects for plants) was the best (lower values for DIC) in describing the SEV and EFA in all the locations being considered. Such a model suggests that the FDR 5788, CDC312 and CDC56 clones displayed greater resistance, while the FX 3864 clone displayed the highest susceptibility in all considered locations.
Key words:
Hevea brasiliensis; Microcyclus ulei; Longitudinal data; Statistical Inference; Stochastic simulation
