Abstract:
The purpose of this thesis was to investigate the use of weights in least squares regression volume table construction and thereby to determine the importance of the assumption of homogeniety of tree volume variance. Several weighted and unweighted linear regression equations were investigated using data from 340 Douglas-fir Pseudotsuga menziesii (Mirb. ) Franca trees from the interior of British Columbia, Canada. The results of the analyses showed that: a) the variance of tree volume for large trees is up to 50 times greater than the variance for small trees; b) the variance of tree volume is directly related to the square of the quantity FD²H and also the square of the quantity D²H; c) erroneous statistical conclusions may be reached if statistical tests are carried out for unweighted regression solutions for tree volume because of the large differences in volume variance; d) the use of weighted least squares regression analysis improved the volume estimate precision for all equation forms tested.
