. 2016 Jun 9;212(1):208–219. doi: 10.1111/nph.14038

Table 2.

Results from multiple Mantel tests in which models differed based on the response variable

Model	Elevation	PBA	Chromene	Dimeric chromane	Location	Overall model
A – all PCs	pMc = 0.007; P = 0.14	pMc = 16.89; P < 0.001^*	pMc = 3.74; P = 0.54	pMc = −13.42; P = 0.18	pMc = −1.69; P = 0.79	R ² = 0.05; P = 0.001^*
B – PC1 and PC2	pMc = 0.01; P = 0.05^*	pMc = 26.94; P < 0.001^*	pMc = −0.06; P = 0.99	pMc = −12.64; P = 0.38	pMc = −1.37; P = 0.88	R ² = 0.07; P < 0.001^*
C – only PC2	pMc = 0.02; P < 0.001^*	pMc = 20.21; P < 0.001^*	pMc = −3.54; P = 0.31	pMc = −4.6; P = 0.43	pMc = −9.82; P = 0.01^*	R ² = 0.26; P < 0.0001^*

pMc, partial Mantel coefficient.

Predictor variables for all models were elevation, prenylated benzoic acid (PBA), chromene, dimeric chromane and GPS location. For the response variable, genetic variation was estimated using the principal component (PC) scores of multi‐locus genotype likelihoods of E. encina individuals transformed into distance matrices. Three matrices were created based on: A, all the PCs; B, PC1 and 2; C, only PC2 scores. Elevation and PBA were significant predictors of genetic variation in Eois encina populations.

*Predictor variables and overall models that were significant.