Abstract
This article proposes a mixed model approach as an alternative to the traditional parameter prediction method in diameter distribution modeling. Unthinned and thinned plots established in mixed stands were used for calibrating a generalized diameter distribution model. The model was based on a two-parameter Weibull cumulative density function (cdf). This cdf was linearized through a complementary log–log link function. Plot random effects were included to account for autocorrelation and the dependent variable, i.e., cumulative stem frequency, was assumed to follow a binomial distribution. With respect to the parameter prediction method, this mixed model approach may generate a consistent estimator with consistent and unbiased variance for the vector of parameters when the variance-covariance matrix of the error terms is properly parameterized. Moreover, the approach enables a better assessment of the different variance components. For the whole group of plots, it provides a predicted average diameter distribution. At the plot level, the random effects can be considered as a departure from this average distribution. As long as the diameter distributions of the individual plots do not exhibit major departures from unimodality, the method proposed in this study should be used to calibrate generalized diameter distribution models.