Also used were the inferential methods canonical correlations analysis (CAP-CCorA) and distance-based redundancy
analysis (dbRDA and DISTLM), routines in the suite of programs for multivariate ecological data developed by M.J. Anderson and available as an add-on (PERMANOVAþ) to PRIMER v6 (Clarke & Gorley 2006; Anderson et al. 2008). Unlike classical canonical correlations analysis, where a multivariate normal distribution has to be assumed, CAP-CCorA, like nMDS, operates on realistic ecologically-based resemblance matrices such as those defined using the BrayeCurtis measure. The question thatwas testedwaswhether therewas an association between the species lists obtained from the subplots (using the ‘species by subplots’ matrix where visits are ignored) and the living vascular plant species present in each subplot. This was achieved
using a two-stage procedure. In the first stage, permutation
tests were performed using CAP-CCorA employing the
two data matrices (one containing the resemblances among
the species lists in each of the 100 subplots and the other
containing the abundances of the vascular plant species
present in those same subplots). Significance levels were
provided by permutation trials, using 9 999 random permutations
of the observed data set. If a significant correlation was
obtained, then DISTLM modelling was used to establish the set
of vascular species that best explained the differences among
the macrofungal species assemblages at the subplot level. In
this stage, the abundances of the 21 vascular plant species (see
Gates 2009, Table 2.2) were transformed using log (Xþ 1) to