Table 2 . - -Value of variables for predicting parameters of trend 



of specific gravity with distance from apex 



Variable ' F-statistic [ Degrees of freedom 



Crown class 17.0** 3,102 



Crown length 16.6** 3,102 



D.b.h. 14.3** 3,102 



Age 2.4 3,102 



Independent of crown length: 



Crown class 6.24** 3,101 



D.b.h. 1.60 3,101 



Significant at the 1% level. 



The 20-year sheath of increment studied here was laid down in the years 1940 to 1959 

 inclusive. Application of the predictors derived below should be restricted to trees subject to 

 moisture regimes similar to those experienced in the Inland Empire during that period. 



The analysis of the previous section showed that this curve changed slope with crown 

 length and crown class, and particular attention was paid to exploring the possible functional 

 form in which these variables might enter the prediction equation for increment specific 

 gravity. Tree height, (H), as a measure of the distance from the apex at which the specific - 

 gravity sample was taken, was also expected to be an important variable. In addition, age, 

 volume growth percent, d.b.h., and diameter growth rate were also considered because other 

 workers had hypothesized that these variables are related to specific gravity. Three separate 

 regressions were computed using 18 to 28 variables representing various transformations and 

 interactions of the variables in the above list. In each regression, two terms were sufficient 

 to explain all but an insignificant portion of the explainable variance.^ 



Breast-height core specific gravity (SC^q) was the most important factor. Its reciprocal 

 was consistently the best functional form of this factor. The best prediction equation of the set 

 solved was: 



^20 " 0-5413 -0.08838 (1/SC^^) +0.9907 (1/H). 



The coefficient of determination was R =0.771 and the standard error of estimate was 

 ±0.0146. 



In this paper, explainable variance is taken to be (Variance of Y) (Max (R^)) where the 

 Max (R ) is the coefficient of determination for the multiple regression on the entire set of 

 independent variables . 



7 



