2012 Volume 17 Issue 1 Pages 1-5
We estimated the occurrence probability of Quercus hondae, an endangered tree species, in understory of sugi plantations by the logistic regression analysis. The occurrence probability was estimated to be high in the older stand situated in compartments including the lower slope. The threshold probability for the best prediction (the cutoff value) was determined as 17 % from the ROC-curve. With this threshold, the model exhibited its high sensitivity (0.90), specificity (0.73), and total accuracy (0.78), indicating its successful estimation of the occurrence probability of Q. hondae.