The unhook chokers model was found to be best explained by its mean: 



UC = 0.88, with a variance of 0.323 



as shown below: 



Mean 

 Median 

 Variance 

 Skewness 



0.88 



.70 



.323 

 2.8166. 



None of the independent variables were found to be significant predictors of this 

 component, probably because of the highly variable conditions and locations of the log 

 decks . 



An alternate method of predicting total turn time (TT') was formed by summing the 

 component models: 



The "best" component models for TU, TLO, and TL were used to predict the time required 

 for these factors. Time required for HC and UC was estimated using their mean values. 

 The correlation between the summed component predictors and the observed data was 

 calculated and then squared. This psuedo R-squared was 0.4264; the standard error of 

 the predicted values was 1.829. These statistics for TT' were then compared with TT 

 model number one (fig. 4) which had an R-squared of 0.4269 and standard error of 1.836. 

 Since the additive model built from the components was more complex than the TT model, 

 it was concluded that the TT model was the best total turn time model for this study. 

 The similarity of the two sets of statistical values indicates that the models were 

 properly selected for both the components and the total turn. 



As a final check, the TT model was tested using 15 observations which had been 

 randomly selected prior to the analysis and set aside for this purpose. The psuedo R- 

 squared (defined in the same way as in the preceding paragraph) was 0.4262, and the 

 standard error was 1.837. These agree very well with the statistics for the best TT 

 model. This implies that the model is doing a reasonable job of predicting the response 

 variable, total turn time. 



The variables in the models developed in this analysis do not explain as much as 

 the variation in the dependent variables as was initially expected. The R-squared 

 values for the dependent variables were: 42.69 percent for TT, 61.68 percent for TU, 

 58.44 percent for TLO, and 27.86 percent for TL. The authors feel that a more quanti- 

 tative method of rating the coded independent variables would increase the R-squared 

 values. This conclusion was based on observations made over a wide range of conditions 

 as compared to the conditions encountered during the relatively short stay of the time 

 study crew. The models also could be improved by expanding the range of the independent 

 variables. This would require many more observations, however, because of the number 

 of combinations of variables and the possibility of interactions between the variables. 



There are other possible sources of error in the equations which would be obvious 

 to anyone who has worked or studied logging operations, but those presented are felt 

 to be most important. 



CONCLUSIONS 



TT ' = TU + TLO + HC + TL + UC. 



13 



