An Estimator of Prediction Error Variance for Projection Equations

Publisher： Society of American Foresters

ISSN： 0015-749X

Source： Forest Science, Vol.54, Iss.6, 2008-12, pp. : 569-578

Disclaimer: Any content in publications that violate the sovereignty, the constitution or regulations of the PRC is not accepted or approved by CNPIEC.

Previous Menu Next

Abstract

An estimator of prediction error variance for projection equations was derived using the first-order Taylor expansion in this study. The estimator, a modified estimator of the prediction error variance for a population mean regression model, was adapted for situations in which projection equations are applied to unsampled individuals. The estimator accounted for the errors associated with the response variable on the right side of a projection equation, as well as the errors associated with parameter estimation and serial correlations in data. The application of the estimator was demonstrated using the example of 140 trees with diameters measured annually from age 5 to 19. The results indicated that the estimator represented prediction error variance well and was useful for constructing prediction intervals for a given projection equation and providing information on the contributions of the variance components associated with different error sources. It was evident that the error associated with the prior observation of the response variable of a projection equation had an important impact on model predictions for forward projections. Overlooking this variance component may result in significant misestimation for the prediction errors of the projection equation.