Four of the best TU models (fig. 5) were examined. In each case, none of the 

 models violated the normality assumptions. Again the differences between models were 

 slight. Model number one was chosen as the "best" model since it was an easy model to 

 use and had a relatively high R- squared and F value: 



TU 



- 7.60 + 0.01173DI - 0.01958DI-SLO + 16.01SLO 



with an R-squared of 61.68 percent. In this model, distance, slope, and distance 

 times slope were important regression predictors of travel unloaded, and these factors 

 are reasonable from a physical standpoint. Other factors such as number of logs and 

 weight were eliminated in the regression analysis, as one would expect, since the carriage 

 is not loaded during the TU element. 



Model #1 



TU = - 7.60 + 0.01173DI - 0. 01958DI -SL0 + 16.01SLO 



Variable 



Cumulative 

 R-squared (%) 



DI 



DI-SLO 



SLO 



44.00 

 52.48 

 61.68 



R 2 = 61.68%, a 2 = 0.37, F = 83.17, 4(x)F- 05 = 10.72 

 Model #2 



TU = 11.21 + 0.00144DI - 36.37SL0 + 31.47SL0 2 



Variable 



Cumulative 

 R-squared (%) 



DI 



SLO 



SLO 2 



44.00 

 47.52 

 51.57 



R 2 = 51.57%, a 2 = 0.47, F = 55.01, 4(x)F- 05 = 10.72 

 Model #3 



TU = 12.80 + 0.00150DI - 39.65SL0 - 0.0121TEMP + 34.99SL0 2 



Variable 



Cumulative 

 R-squared (%) 



DI 

 SLO 

 TEMP 

 SLO 2 



44.00 

 47.52 

 51.57 

 54.32 



Increase in 

 R-squared (%) 



44.00 

 8.48 

 9.19 



Increase in 

 R-squared (%) 



44.00 

 3.52 

 4.05 



Increase in 

 R-squared (%) 



44.00 

 3.52 

 4.05 

 2.75 



R 2 = 54.32%, o 2 = 0.45, F = 45.77, 4(x)F- 05 = 9.8 

 Model #4 



TU = - 8.79 + 0.01425DI - . 02096DI • SLO + 17.98SL0 - . 000022DI -TEMP 



Variable 



Cumulative 

 R-squared (%) 



Increase in 

 R-squared (%) 



67.94% 



DI 



DI-SLO 

 SLO 



DI -TEMP 

 J 2 = 0.32, F = 



44.00 

 52.48 

 61.68 

 67.94 



SI . 60, 4(x)F- 05 = 9.8 



44.00 

 8.48 

 9.19 

 6.26 



Figure 5. --Travel unloaded (TU) statistics 



9 



