Permanent sample plots have become the main source of information for estimating models which quantify the dynamic processes of a forest. Fitted models allow for projecting inventories, used to determine timber production and many forest management decisions. The quality of these models is largely dependent on the quality of the information provided by the permanent sample plots. However, the pool of information contained in recent permanent sample plots is limited. Efficient estimation techniques must use all the information available from such plots. Current estimation techniques can be improved. Existing techniques employed in forestry have failed to recognize the random nature of the individual model characterizing each plot. On the other hand, techniques designed for remeasured entities in other scientific fields do not address particular forestry situations such as the small number of remeasurements or the irregularity of remeasurements. A framework for estimating forestry growth models which recognizes the individuality of each plot and special forestry situations is presented in this dissertation. The proposed framework is a two-stage estimation technique, in which the growth rate of a permanent sample plot is considered analogous to the interest rate on a bank account. The first stage estimates the growth rate after removing the time effect. The second stage, based on Von Bertalanffy's growth curve, relates growth rate to site index and volume at the beginning of the growing season. The proposed predictor of future growth rates, the "weighted predictor," is a weighted average between the growth rate observed on a plot and the growth rate predicted from the second-stage model. The weighted predictor is then used to compound the current volume of a plot. An estimate of the variance of the prediction can also be computed. / Forestry, Faculty of / Graduate