388 



University of California Publications in Geological Sciences 



Applications and Reliability 



The superb preservation of Rancho La Brea material has allowed the application 

 of multivariate analysis to samples large enough to have statistical significance. But 

 often only small samples can be assembled, and it seems reasonable that some sort 

 of multivariate analysis similar to that used in this study could be applied to such 

 samples, with emphasis placed on the descriptive value of the technique. The multi- 

 variate test can provide an evaluation of subtle quantitative differences when the 

 interpretation of the differences between single measurements on two or more 

 samples is difficult. 



TABLE 6 



Percentage Comparisons of Distance (D) by Different Samples 



Samples compared 



Cranium 



Mandible 



ochropus-lestes /ochropus-mearnsi 



62.5 

 74.1 

 84.4 

 65.6 

 56.6 

 41.0 

 42.8 

 86.3 

 95.9 

 47.8 

 57.8 

 82.7 



60.5 



ochropus-lestes/mearnsi-lestes 



mearnsi-lestes /ochropus-mearnsi 



ochropus-mearnsi/ ochropus-orcutti 



ochropus-mearnsi/mearnsi-orcutti 



ochropus-lestes / ochropus-orcutti 



ochropus-lestes/lestes-orcutti 



ochropus-orcutti/ mearnsi-orcutti 



lestes-orcutti/ ochropus-orcutti 



lestes-mearnsi/mearnsi-orcutti 



lestes-mearnsi/lestes-orcutti 



lestes-orcutti /mearnsi-orcutti 



69.3 

 87.2 

 40.6 

 36.1 

 24.5 

 28.1 

 89.1 

 87.3 

 31.5 

 40.5 

 77.7 







Fragmentary material creates another problem in paleontological work: the 

 measurements used here for the study of the coyotes might be impossible to obtain 

 from other samples because the material was distorted or incomplete. In studies of 

 such material, measurements of several small but taxonomically important char- 

 acters might be useful. In table 6 the distance (D) between each group and an- 

 other (given in figures 3 and 4) has been expressed as a percentage of the 

 distance between one of the pair and a third group. Since the percentages based 

 on the eight-variate cranial analysis are, with only two exceptions, larger than 

 those based on the five-variate mandibular analysis, the cranial analysis has 

 separated the groups to a greater degree. But with the perfect Recent material 

 (the first three entries in table 6), the five mandibular measurements provide a 

 basis for classification that is practically as good as that provided by the eight 

 measurements of the cranium. Imperfections in the Rancho La Brea material 

 result in the five-variate analysis having the same definite trend but not the same 

 absolute magnitude as the eight-variate procedure. 



