An improved estimator for the sampling error of local competition variables

Abstract

We present a revised estimator for the sampling error of local competition variables that builds on the conceptual framework given by Stage and Wykoff (Stage, A.R., and Wykoff, W.R. For. Sci. 44(2): 224–238, 1998). Accurate estimation of the sampling error of local competition variables is a requisite for most approaches that correct the effects of measurement error in model fitting and application. Our revision addresses the bias inherent in Stage and Wykoff's estimator due to the overlapping of random samples that are constrained to include a subject tree. We also argue that the adjustment that Stage and Wykoff used to account for the absence of treeless plots (zero truncation) is unnecessary. We illustrate the performance of the new estimator and the estimator of Stage and Wykoff through simulation. For a hypothetical Poisson forest (800 trees·ha−1; mean diameter at breast height, 9.8 cm), bias is negligible for the new estimator, and variance is reduced by 92%.