We applied the 1.45 circuity factor to all of the observations in our study. We then estimated using generalized least squares (i.e., allowing a heteroscedastic covariance matrix) a quadratic model that related stand age as measured in FIA's sixth inventory (1990 in North Carolina, 1992 in Virginia) to distances to nearest mill (miles), distance to nearest road (miles) as recorded by FIA, site slope (percent), and site quality as recorded by FIA (varying in our sample from 1 to 5, where 1 is the highest productivity, 5 is lowest).
In a Box-Cox approach, where parsimony is perceived as an important model selection criterion, variables that had t-values less than unity in absolute value in a first round of estimates were dropped for a second round. The second round produced a parsimonious estimate. We report in the following slides relationships between stand age and these variables corresponding to the final, parsimonious models.
