The Use of Bayes/Empirical Bayes Estimation in Individual Tree Volume Equation Development

Two empirical Bayes estimators were presented for the parameters in the combined variable volume equation. For the slope coefficient, an informative prior distribution was used, while noninformative priors were specified for the error term and the intercept. The empirical Bayes estimators were compared to weighted least squares estimators for loblolly pine, white oak, black cherry, and red maple through simulation studies. The empirical Bayes estimators were superior in terms of predictive ability for white oak and black cherry. There was little difference between empirical Bayes and least squares for loblolly pine, and least squares was superior for red maple. These results were attributed to the quality of the prior information for each species. Finally, there was an indication that in cases where good prior information exists, empirical Bayes estimation may be used to reduce the amount of data necessary to construct volume equations. Forest Sci. 31:975-990.