Comparing Linear Discriminant Analysis with Classification Trees Using Forest Landowner Survey Data as a Case Study with Considerations for Optimal Biorefinery Siting

Bioenergy is reemerging as an important topic in energy‐related research. The rapid increase in costs of petroleum products has led to a renewed interest in alternative sources of energy such as biofuels. World‐wide energy consumption has increased 17 times in the last century and the demand for energy in emerging markets such as China and India is projected to increase in the future at unprecedented rates. A review of the current bioenergy literature is presented in this thesis. Also, comments on the economics of bioethanol are discussed. The primary part of the thesis focuses on statistical classification methods related to factors that influence landowner attitudes towards harvesting timber. A comparison of linear discriminant analysis (LDA) and classification tree (CT) methods is presented using the results of a forest landowner survey as a case study. Several CT techniques are discussed with an emphasis on the CRUISE classification tree program. The LDA procedure in SPSS is used to construct linear discriminant functions of the survey results. CRUISE is also used to construct classification trees of the survey results. Survey results showed that 73.3 percent of farmer forest landowners harvested timber, and 69.6 percent of non‐farmers who had a length of residency beyond 36.5 years harvested timber. For landowners who conducted commercial timber harvests, the importance level of income from the harvest was the overriding factor relative to all other factors. Discriminant analysis results supported the results of CTs. However, the linear discrimination functions and corresponding coefficients did not provide the level of two‐dimensional detail of CTs, which also detected hidden interactions