OTUs related to host performance
OTUs of differential abundance between ranges and temperatures were identified from the univariate output of the mGLM fitted on post-disturbance time points. We considered OTUs with p-values < 0.05 and with coefficients of which the 95% confidence limits were either both positive or both negative as differentially abundant. Following a joint modeling approach (Warton et al., 2015), we used the residuals of the mGLM to calculate Spearman correlations coefficients between OTUs in the samples of the final timepoint (t12) and the RGR and observed thallus brittleness disease symptom, for which 95% confidence intervals were obtained by bootstrapping with a 1000 iterations.
The univariate GLMMs for host performance traits and diversity measures were fitted using the R package lme4 (Bates, Mächler, Bolker, & Walker, 2015). nMDS was conducted using the R package vegan (Oksanen et al., 2017). The mGLMs were ran using the R package mvabund (Y. I. Wang et al., 2012). To consider time in a flexible way in all post-disturbance models, we compared different versions of the full model based on the AICc or AICsum (for the mGLM), where time was specified as factorial and continuous variable (linear, log transformed, squared root transformed and as polynomial). Violations of model assumptions were verified visually with QQ-plots and residual vs fitted-plots for univariate and multivariate analyses.