Table 3.— The prediction model for diameter growth of nonreleased western redcedar 



Nonreleased diameter growth prediction model 



PGRTH = B(0) + B(1) DIA + B(2) ELEV + B(3) COSINE + B(4) SINE + B(5) SLOPE + B(6) THATFI + B(7) THPAMY + B(8) TSPAMY 



where: 



PGRTH = predicted nonreleased diameter growth, 



DIA = natural log d.i.b. at beginning of growth period in mm 



ELEV = stand elevation above sea level in feet divided by 100 



COSINE = cosine of aspect azimuth in degrees times the slope in percent 



SINE = sine of aspect azimuth in degrees times the slope in percent 



SLOPE = slope in percent 



THATFI = 1 if THPUATFI habitat type, otherwise 



THPAMY = 1 if THPUPAMY habitat type, otherwise 



TSPAMY = 1 if TSHE/PAMY habitat type, otherwise 



B(0) = 3.496337 

 B(1) = 1.120681 

 B(2) = -0.018654 



R 2 = 0.69 



B(3) = 0.007143 

 B(4) = - 0.000368 NS 

 B(5) = -0.022533 



B(6) = 0.596686 

 B(7) = 0.538276 

 B(8) = 0.251484 NS 1 



Independent variables tested: 



1. Stand location — geographic 



2. Habitat type — TSHE/PAMY, THPUPAMY, ABLA/PAMY, THPUATFI 



3. Stand elevation above sea level — ft 



4. Slope configuration 



5. Cosine of the aspect azimuth in degrees times percent slope 



6. Sine of the aspect azimuth in degrees times percent slope 



7. Natural log diameter inside bark (d.i.b.) at beginning of growth period in mm 



8. Breast high tree age at release in years 



9. Slope in percent 



Dependent variable: DDS 



1 NS = nonsignificant at P < 0.05, the variable was included because other variables of the same type were significant. 



A natural log transformation of the difference 

 between the squared diameters (DDS) was used to 

 linearize the diameter-growth measurements as follows: 

 [In (D^ - Djj)] where: D, equals the d.i.b. at the begin- 

 ning of a growth period and D 2 equals the d.i.b. at the 

 end of a growth period. Examinations of plots of the 

 data and other diameter-growth models indicated a log- 

 normal distribution for the squared differences 

 (Wykoff 1 ). Using the transformed diameter growth for a 

 5-year period before treatment as the dependent 

 variable, a nonreleased diameter-growth model was 

 fitted, using all possible regressions for the largest R 2 , 

 which in turn was evaluated stepwise for the significant 

 (P<0.05) independent variables. The nonreleased 

 model is a reliable predictor of diameter growth, 

 explaining 69 percent of the variation (table 3). 



Wykoff, W. R. Personal communication. Intermountain Forest and 



Range Experiment Station. Moscow. Idaho. 



IDENTIFYING SIGNIFICANT ASSOCIATIONS 



Regression analysis was used to identify the site, 

 stand, and tree characteristics that were associated 

 with the diameter-growth response of western redcedar. 

 The dependent variable was the difference between 

 observed diameter growth and the predicted non- 

 released diameter growth for a tree through the same 

 time period. Using the observed d.i.b.'s at time of 

 release, 5. 10, and 15 years after release and for the 

 years 1975 and 1979, four transformed 5-year 

 diameter-growth measvirements (DDS), were calculated. 

 Using the observed growth measurements and the 

 predicted growth measurements, the dependent vari- 

 able [DDS "released"] - [DDS "predicted"] was 

 computed. The predicted diameter growth (pred. 

 growth) of each tree was included in each regression 

 model to eliminate variation not related to the release 

 response of the sampled trees. Also, if either the sine- 

 aspect term or cosine-aspect term was significant, both 

 terms remained in the model. The regression models 

 could then be formulated to identify the site, stand, and 

 tree characteristics that were associated (P5 0.05) with 

 the difference between observed diameter growth and 

 predicted nonreleased diameter growth of western 

 redcedar. 



5 



