Decision tree applications for forestry and forest products manufacturers
Young, T. M., Y. J. Wang, D. G. Hodges, and F. M. Guess.  2008.  Proc. of 2008 Southern Forest Economics Workshop. Savannah, GA. Ed. Tommy Tye. Center for Forest Business, University of Georgia, Athens.

Research results are presented for two studies which use the common statistical method of decision trees. First, regression trees of the internal bond (IB) of 0.750¡¨ medium density fiberboard (MDF) identified significant regressors (ƒÑ < 0.05) as ¡§refiner steam pressure¡¨ which interacts with ¡§press start control¡¨ and ¡§dry fuel bin speed.¡¨ The highest IB (mean = 151 psi) occurred with ¡§refiner steam pressure¡¨ >_54.6 and ¡§press start control¡¨ „T _933.0. The lowest IB (mean = 132 psi) occurred with ¡§refiner steam pressure¡¨ „T_54.6 and ¡§dry fuel bin speed¡¨ „T_27.7. Second, classification trees of 495 forest landowners in the Cumberland Plateau region of Tennessee revealed that the most significant factor (ƒÑ < 0.05) influencing tree harvests was whether or not the respondent was a farmer. Seventy-three percent of farmers had harvested timber previously. For those who were not farmers, the most significant factor (ƒÑ < 0.05) was ¡§years residing at current address.¡¨ If a non-farmer resided at the current resident longer than 36.5 years the chance of harvesting timber was 69.6 percent. For landowners that conducted commercial timber harvests, the only significant factor (ƒÑ < 0.05) from the classification tree was unsurprisingly the importance of income from the harvest. Seventy-five percent of these respondents conducted a commercial timber harvest.