Gates 2 and 3 



At gates 2 and 3, the road network, size and number of 

 cutting units, yarding system, silvicultural methods, and 

 harvested acres are defined. These are examples of vari- 

 ables used to develop the gate 2-3 equations. At this 

 point, the auction date is approximately 1 to 3 years in 

 the future. 



EASTSroE EQUATIONS 



The equations and classification results are presented 

 in table 5. The gate 2-3 eastside equations have the same 

 significant variables and differ only in the transforma- 

 tions used. The equations indicate the more volume har- 

 vested (TOTVOL), the more likely the timber sale will be 

 sold. The standardized discriminant coefficient for total 

 volume harvested indicates it is the most important deter- 

 minant of sold and unsold sales. If the average logs per 

 thousand board feet (ALPM) is high, it is more likely that 

 the sale will be unsold. The ALPM variable is the number 

 of logs to be moved; the more pieces moved the less likely 

 the sale will be sold. The final variable — ^miles of total 

 road construction (TOTROAD) — ^represents the initial 

 development necessary to harvest the sale. The miles of 

 road construction affect the sale by restricting the number 

 of potential purchasers and introduce an additional source 

 of risk by delaying the harvest (Johnson 1979). 



The classification results for the two methods are al- 

 most identical (see table 5b). The difference lies in the 

 number of sold sales correctly predicted. The logistic 

 regression correctly classifies an additional sale. The 

 discriminant fiinction classification results are very 

 stable. The classification results are identical for both 

 nieasures of prediction accuracy. 



The TOTVOL variable is the only gate 1 variable re- 

 maining. TOTSALE has been displaced by gate 2-3 vari- 

 ables (ALPM and TOTROAD). ALPM and TOTROAD are 

 more specific timber sale information and allow prediction 

 accuracy to increase. 



Using gates 2 and 3 information, the equations have 

 improved the overall classification results by 23.1 percent 

 for the logistic regression and 6.9 percent for the discrimi- 

 nant equation. 



WESTSroE EQUATIONS 



The westside gate 2-3 equations are quite similar in 

 terms of significant variables. Total road construction 

 (TOTROAD) was a more desirable variable in the logistic 

 regression than was miles of new road construction 

 (NEW) and miles of road reconstruction (RECON). 



The equations and classification results are presented 

 in table 6. The following variables have a positive effect 

 on a sale selling: the ratio of acres harvested to total sale 

 acres (DENSE), volume-per-acre harvested (VPA), per- 

 centage of volume tractor yarded (%TRAC), and the aver- 

 age maximum tractor yarding distance (TRACDIST). The 

 remaining significant variables, miles of road construction 

 (TOTROAD, NEW and RECON), percentage of volume of 

 dead lodgepole or dead whitepine (DEAD), and the aver- 

 age logs per thousand board feet (ALPM) have a negative 



Table 5 — Gate 2-3 eastside equations and classification results 



A. Equations 





Logistic regression 



Discriminant analysis 



Variabis 



Coefficient fStd Err) 



Coefficient 



(Std CoefO 



Ln(TOTVOL) 



3.812 (1.448) 







(TOTVOL)"^ 





0.073 



(1.537) 



Ln(ALPM) 





-.962 



(.487) 



ALPM 



-.042 (.060) 







(TOTROAD)"' 



-3.440 (1.342) 



-1.105 



(-1.294) 



Constant 



-18.426 (7.699) 



2.142 





B. Classification results 





Logistic regression 



Discriminant analysis 



Actual Total Correct Percent 



Correct 



Percent 



group sales predict correct 



predict 



correct 



Unsold 



14 10 71.4 



10 



71.4 (71.4)' 



Sold 



26 22 84.6 



21 



80.8 (80.8)' 



All sales 



40 32 80.0 



31 



77.5 (77.5)' 



'Indicates percent correctly classified using the holdout method. 



Table 6 — Gate 2-3 westside equations and classification results 



A. Equations 



Logistic regression 



Discriminant analysis 



Variable Coefficient 



(Std Err) 



Coefficient 



(Std Coeff) 



DENSE 



0.864 



(0.420) 







(DENSE)"' 







1.344 



(0.324) 



VPA 



.062 



(.016) 







(VPA)'" 







.489 



(.651) 



Ln(TOTROAD) 



-.200 



(.112) 







(NEW)'" 







-.230 



(-.301) 



(RECON)'" 







-.064 



(-.082) 



DEAD 



-.026 



(.010) 







Ln(DEAD) 







-3.944 



(-.373) 



(ALPM)' 



-.0004 



(.00030) 







(ALPM)'" 







-.261 



(-.269) 



%TRAC 



.022 



(.004) 



.022 



(.778) 



(TRACDIST)' 



.00002 



(.0000) 







Constant 



•1.300 



(.450) 



-2.327 





B. Classification results 





Logistic regression 



Discriminant analysis 



Actual Total Correct Percent 



Correct 



Percent 



group sales predict correct 



predict 



correct 



Unsold 145 



84 



57.9 



101 



69.7 (66.9)' 



Sold 204 



161 



78.9 



139 



68.1 (66.2)' 



All sales 349 



245 



70.2 



240 



68.8 (66.5)' 



'Indicates percent correctly classified using the holdout method. 



6 



