SAMPLING ASSUMPTIONS OF 



ALGORITHM 



When data have been collected according to a sampling design, each observation 

 represents a number of similar observations. When this number (weight or blowup 

 factor) varies from observation to observation (i.e., when sampling is other than equal 

 probability sampling), the effect on the relationship of interest must be recognized, 

 particularly in estimating individual tree mortality. 



Because of the distribution of mortality in endemic situations, wide areas must 

 be sampled to assure an adequate representation of dead trees. It would not be 

 feasible to measure each green tree in the area; therefore, only a subsample of the 

 green trees is measured. In this situation not all trees are sampled with equal 

 probability. For example, assume 5-acre plots have been selected for obtaining mor- 

 tality counts. A 1/5-acre plot is established within each 5-acre mortality plot to 

 obtain counts of green trees. Thus, dead trees have a relative weight of 1 and green 

 trees have a relative weight of 5/(1/5) = 25. 



5 



