Modeling dominant height through site index, defined as dominant height at
some fixed tree age, requires two equations. The site index must be estimat
ed as the average height of dominant trees at base age, and dominant height
must be modeled as a function of age and site index, each model having its
own error term, This raises a problem when squared errors are minimized to
estimate parameters of interest: the estimates are biased and inconsistent
because the error terms are correlated with the explanatory variables. Two
additional problems are the autocorrelation among error terms and variance
heterogeneity within trees. The generalized method of moments (GMM), which
simultaneously takes these problems into account, offers a tool to estimat
e parameters of the site index and dominant height equations consistently a
nd without bias. Allowing parameters of the height growth model to depend o
n explanatory variables representing ecological region and drainage class p
rovides useful insight into growth patterns, and improves prediction based
on these models. A simulation study confirms that the GMM combined with pre
whitening of the time series for each tree produces useful estimates of the
parameters and valid statistical tests of hypotheses about their value.