Empirical Bayes Development of Honduran Pine Yield Models

Publisher Website
Google Scholar
Abstract
Occasionally it is of interest to calibrate a given growth or yield model to data from several regions. If it is expected that the parameters from different regions can somehow be regarded as similar, then a Bayesian approach suggests itself. A cursory examination of the literature reveals that most of the theoretical work on empirical Bayes estimation for the linear model has focused on simultaneously estimating the coefficients in one model. In these methods the usual least squares estimates of the model parameters are shrunk toward their mean. One set of parameter estimates results. In contrast we desire a method whereby multiple sets of parameter estimates are produced. We report the results of using such a method to calibrate yield models for unthinned Honduran pine plantations to data from 21 soil-site groups. The models compare favorably to those developed via traditional methods and allow estimation of regression coefficients even for soil-site groups for which the design matrix is not of full rank. For. Sci. 38(1):21-33.

This publication has 0 references indexed in Scilit: