Principal Component Analysis
To examine the association between plant infochemicals and host location and selection in closely related host-flexible parasitoids a principal component analysis was performed using the relative area of the chromatogram peaks, a measurement of abundance for each compound, obtained from the GC/MS and parasitoid-host association data. Four separate principal component analyses were run in R using the package “FactoMineR” (Lê, Josse, & Husson, 2008) and visualized using the package “Factoextra” (Kassambara & Mundt, 2017). The first principal component analysis included all Eucalyptus leaves collected to examine differences in phytochemistry between the two species ofEucalyptus sampled (globulus vs. sp. 1). The second principal component analysis included damaged and undamaged flushEuc. globulus leaves(the age class preferred by the beetles) to explore variation in plant chemoprofiles when damaged, suggesting a specific herbivore-induced response. Only Euc . globuluswere examined because there were not enough samples to adequately assess herbivore-induced responses in the other Eucalyptus species. The third principal component analysis included damaged leaves corresponding to unparasitized beetles, with beetles identified by their distinctive larval coloration: yellow body with a prominent black dorsal strip inParopsisterna cloelia ; and black body with orange/yellow lateral stripes in Pst. agricola . This subset of samples permitted examination of the relationships between infochemicals and the observed herbivore assemblage. The fourth principal component analysis included damaged leaves corresponding to parasitized beetles from which different species of Eadya were reared, in order to examine the relationship between infochemicals from the plant and Eadya host selection To minimize the effects of pseudo-replication in the principal component analysis in the parasitoids and beetles, the dataset was collapsed to only include one sample per Eadya species for each tree. To confirm separation in each of the four principal component analyses, one-way ANOVAs were performed using the first and second principal components, with Tukey post-hoc tests for the parasitoids and beetle analyses. To determine phytochemical compounds which may be utilized as infochemicals by Eadya , compounds were ranked in descending order by the absolute value of PC1 – PC2 variable loadings. The R script for the PCA and ANOVAs, as well as all data files, can be found on Figshare (www.figshare.com10.6084/m9.figshare.17105768).