A group contains one or more sets. The five groups in these screening runs were established 

 on the following criteria: group 1 contained the single variable, ln(D); group 2 consisted of 

 the seven powers on D; group 3 was the two rainfall variables; group 4 was the total stand 

 density variables; and group 5 was the position in diameter class sets. Site index was not 

 included in these runs because the difference between the site index values on each half plot 

 was probably too small to prove significant in modeling. Program REX was then set up to 

 tabulate the RMSQR values of all regression equations formed by choosing, at most, one set 

 from each group. 



From these screening runs, 12 blackjack and yellow pine equations were selected and the 

 regression coefficients determined. These coefficients were next examined for reasonableness 

 of behavior and for consistency of performance between plots. In addition, the model with the 

 lowest RMSQR was picked for each plot and the residuals of these models were checked for 

 normality. To do this, skewness and kurtosis statistics were computed using formulas found in 

 Kendall and Stuart (1977) . From this examination, I concluded that the residuals are not 

 normally distributed, that they are highly skewed, and are leptokurtic (Kendall and Stuart 

 1977) . One consequence of the lack of normality is that all of the usual significance tests 

 that could be made on regression results are not applicable. 



Another problem that made testing difficult or impossible was the lack of data range and 

 overlap between the different stand conditions. The site index of the even-aged plots was 

 larger than the uneven-aged plots; the stand structures were obviously different; and the even- 

 aged plots covered only a narrow range of diameter classes. For the uneven-aged stands, the 

 managed and virgin plots differed in stand structure due to cutting. If legitimate tests were 

 available, a test indicating significant differences in the models of each stand condition 

 might be due to the data being truly incompatible, or merely that the the data are complementary. 

 Because of these problems, I concluded that testing to see if individual data sets could be 

 pooled was not practical. 



The main question was whether the even-aged and uneven-aged data sets could be combined 



for blackjack pine. To answer this without testing, it was decided to develop two blackjack 



pine equations: one with the even-aged data, and one without. Both of these could then be 

 evaluated as to their predictive capability. 



A second set of screening runs was then made with the plots combined as such: for both 

 blackjack pine and yellow pine, the virgin, uneven-aged plots (61 and 62) were combined; and 

 the cut, uneven-aged plots were combined (71 and 72). In addition, another set of blackjack 

 pine data was created by combining the Taylor Woods even-aged data with the cut, uneven-aged 

 data. From the findings of the first screening, the independent variables T, PCT, and DCP 

 (and their transformations) were eliminated. Site index (S) was added as another group. 



The results of this second set of screening runs were examined, and 11 blackjack pine and 

 4 yellow pine equations were selected and the regression coefficients determined. As before, 

 these coefficients were also checked for reasonableness of behavior and for consistency of 

 performance between data sets. For blackjack pine, those independent variables that both 

 behaved reasonably and minimized RMSQR were ln[D) , D, Jn(S), in(GRF), LBA2, MBA2, and UBA2 . 

 For yellow pine, the independent variables included ln{D) , D, in(S), in(GRF), (LBA2)^, (MBA2)^, 

 and (UBA2)^. The independent variable of Jn(S) was inconsistent for blackjack pine and 

 unreasonable for yellow pine. It was included, however, because of the desire to develop a 

 simulator applicable for the range of site indices found in the habitat type. The problems 

 with site index persisted throughout the analysis and finally necessitated special handling, 

 which will be discussed later. 



