Using joint multivariate analyses of leaf morphology and molecular-genetic markers for taxon identification in three hybridizing European white oak species (Quercus spp.)

Table 1 Separation of the clusters (representing oak species) using two multivariate approaches with three variable sets (all, only morphological, only genetic)

Approach	Variable set	n (Pet)	n (Pub)	n (Rob)	Cohesion within clusters	Separation among clusters	Separation/cohesion ratio
FAMD	All variables	466	428	475	1.01	3.87	3.84
	Morphological variables only	431	437	501	1.01	3.47	3.43
	Genetic variables only	510	439	420	1.10	1.19	1.09
LDA	All variables	430	471	468	0.90	2.99	3.33
	Morphological variables only	388	450	531	0.97	2.38	2.45
	Genetic variables only	419	476	474	0.90	2.80	3.13

n = number of trees assigned to the different species (Pet = Quercus petraea, Pub = Q. pubescens, Rob = Q. robur) based on HCPC (hierarchical clustering of principal components) for FAMD or highest probability (majority rule) for LDA. Cohesion represents the weighted average distance within clusters. Separation is the average distance among clusters. A large separation/cohesion ratio indicates good separation among species. Distances are based on the first two axes of the multivariate analyses
FAMD factor analysis of mixed data, LDA linear discriminant analysis

ISSN: 1297-966X