MODELING METHODS 



Data Screening 



The first step is to screen the independent variables for those that best explain 

 variations in the dependent variable (mortality rate] . 



In the model to be fit to the data, the dependent variable is dichotomous: the 

 value is assigned to green trees, the value 1 to dead trees. For model fitting, 

 probability of mortality for an individual tree is not considered; a specific tree is 

 either alive or dead. The analytical method enables us to use all trees with similar 

 characteristics (independent variables) to estimate the probability of mortality for 

 a tree with these characteristics. 



In fitting the model, we assume that each tree is an independent observation. 

 Thus, we ignore the fact that there is some clustering in the selection of sample trees. 

 However, in most instances the model will be applied to stands; thus, the impact of 

 clustering, and of the resulting nonindependence of observations, is reduced. 



The algorithm used in this screening process has been developed specifically for 

 situations in which the dependent variable is dichotomous. SCREEN, the computer program 

 that does the screening, is described by Hamilton and Wendt (1975) . The theoretical 

 development of the algorithm and the statistic used for screening are discussed by 

 Gleser and Collen (1972) and by Sterling and others (1969) . 



4 



