V 



PROPERTIES OF 

 MORTAUTY MODELS 



Models can be developed to predict either stand mortality or the mortality of 

 individual trees within the stand. The models discussed in this paper were developed 

 to predict individual tree mortality within a specified time interval as a function of 

 stand and individual tree characteristics. The true probability of mortality is 

 bounded by and 1. Thus, it would be desirable for the mortality model to -provide 

 estimates of probability in the interval [0,1]. 



A linear regression model frequently has been used for this purpose. However, a 

 linear model cannot insure that probability estimates will be in the interval [0,1]. 

 In addition, using a linear regression model for this purpose creates many statistical 

 problems (Hamilton 1974) . 



Walker and Duncan (1967) and Neter and Maynes (1970) have suggested that the 

 logistic function, which limits estimates of probability to the interval [0,1], is 

 the statistically preferred model for expressing the relationship between a 0,1 depend- 

 ent variable and a set of independent variables. For this reason, we have used a 

 logistic function for modeling mortality. 



