2006 Volume 5 Pages 63-83
In this paper, we presented a statistical procedure of estimating the clustered multivariate linear regression model for predicting the amount of carbon sequestered in a forest stand where there exist several growth patterns. The procedure is as follows: 1) By fitting a volume growth curve to the data of each sampled tree, parameters of the applied growth curve model are estimated. 2) By setting the estimated parameters as new observations, we classify growth patterns of sampled trees by k -means method based on the new observations. 3) We construct a multivariate normal linear regression model with dummy variables for k -clusters. 4) Among a set of the estimated regression models with the different number of clusters, the best model is selected by minimizing the resultant predictive Akaike’s information criterion (PAIC) for the remaining trees. 5) Finally, by using the best set of parameters of growth curves for the remaining trees, we predict the amount of carbon sequestered by remaining trees with its asymptotically 1−α confidence interval.