Comparison of estimation techniques for a forest inventory in which double sampling for stratification is used

Abstract

A number of different estimators can be used when forest inventory plots cover two or more distinctly different condition classes. In this article the properties of two approximate Horvitz- Thompson (HT) estimators, a ratio of means (RM), and a mean of ratios (MR) estimator are explored in the framework of double sampling for stratification. Relevant theoretical properties are given, and a simulation study is used to compare the properties of these estimators. It is concluded that the variance of the RM estimator is smaller than that of the approximate HT for the populations studied, but that of the MR estimator is larger. While ratio estimators are known to be biased for small sample sizes, an additional bias also occurs when the population of interest has an edge-effect. The RM classical variance estimator underestimated the true variance in the simulation study. The approximate HT estimators exhibited biases that were generally smaller than those of the RM and MR estimators for populations with an edge-effect. Its classical variance estimator was essentially unbiased.

Citation

Williams, Michael S. 2001. Comparison of estimation techniques for a forest inventory in which double sampling for stratification is used. Forest Science. 47(4): 563-576.