ABSTRACT
This study was based on the application of the nonlinear multilevel mixed-effects theory in modeling the height growth of Eucalyptus plantation. The database was from individual tree measurements, taken from different sites and over time. This type of database is considered as longitudinal, irregularly spaced, unbalanced, with autocorrelation and heteroscedasticity. The tree-parameter logistic model was used to estimate the height growth with fixed and random effects in two levels: sample units (level 1) and trees inside sample units (level 2). By including both levels, the standard error of estimate was reduced significantly. Also, the estimates were improved by modeling the variance heterogeneity and the autocorrelation, using the ARMA(2,1) structure.
Key words:
multilevel mixed-effects model; Eucalyptus growth; heteroscedasticity; autocorrelation