ABSOLUTE VS. RELATIVIZED COVERAGE DATA FOR CLUSTER ANALYSIS 



In this study, absolute coverage which was expressed as the midpoint 

 values of cover classes was used for determining similarities. Coverage 

 estimates were corrected, insofar as it was possible, at the time of samp- 

 ling to account for recent grazing. 



Gauch and Whittaker (1972) in their comparison cf ordination techniques 

 found that using absolute coverage did not produce different results than 

 using relative coverage, except for principal component analysis. This may 

 not be the case, at least for cluster analysis. This can best be shown 

 using some hypothetical data. 



Percent Absolute Coverage Percent Relative Coverage 



Stand 



Species 



1 

 2 

 3 



4 



Using Sorensen's index and absolute coverage. Stands A and C and Stands 

 A and B have about the same similarity (eighty-eight percent and eighty-seven 

 percent respectively) while Stands B and C d.vQ. 76% similar. Using the same 

 index and relativized coverage. Stands A and B d^r^ ninety-eight percent simi- 

 lar, while Stands A and C and Stands B and C d.vQ. both seventy-nine percent 

 simi lar. 



No doubt some would say that either method is legitimate if consistently 

 applied, but to the ecologist, they do not provide equal information. Rela- 

 tivizing data involves a loss of information--namely the true amount of cov- 

 erage. Presumably some factor, and not necessarily grazing has caused the 

 difference in coverage betv/een Stands A and B, and to say that they are ninety- 

 eight percent similar hides this fact. 



On the other hand, the use of absolute coverage, while more clearly dif- 

 ferentiating between Stands A and B, may obscure the difference between Stands 

 A and C. However, no data have been lost. The ecologist can see that the cal- 

 culated similarities for Stands A and C and Stands A and B are similar but for 

 different reasons. It is then up to the ecologist to determine the signif- 

 icance of the difference in coverage between Stands A and B and also the sig- 

 nificance of Species 4 in Stand C. Perhaps it is an invader reflecting gra- 

 zing or disturbance, or is it a species of narrow ecological amplitude re- 

 flecting a significantly different site and perhaps climax community. This 

 a posteriori weighting seems to be preferable to the de facto a priori weight- 

 ing of species 4 if relativized dates ^.r^ used. 



13 



