The results were summarized using an anova, and detailed post hoc (Tukey) tests were performed with the multcomp package in R. Rabbit pellet data were analysed using the above model with ungrazed plots removed from the analysis. For soil analysis and diversity, the effect of area alone was tested using a GLM, with pairwise comparisons of area?��?treatment undertaken using post hoc Tukey tests (Minitab 13.31; Minitab Ltd, Coventry, UK). The floristic data were analysed using multivariate analyses. Initially each data set was analysed using a detrended correspondence analysis, using Canoco for Windows (version 4.5) package (ter Braak & ?milauer 2002). As the gradient lengths RAD001 manufacturer
were short (<4.0 SD), subsequent analysis was undertaken using linear ordination methods (ter Braak & ?milauer 2002): redundancy analysis (RDA) and principal components analysis (PCA). For all analyses, the species composition data were log-transformed (ln x?+?1) and standardized by sample norm. Effects of grazing on species composition. Partial constrained RDAs were used to test whether the grazing treatments had a significant effect on species composition within each area. Significance was tested using restricted Monte Carlo permutation tests (499 permutations) with an appropriate randomization design. Permutations were restricted for a split-plot design (ter Braak & ?milauer 2002); split plots (treatments) were freely permuted within whole plots (blocks). <a href="http://www.selleckchem.com/products/Rapamycin.html
">www.selleckchem.com The significance of all canonical axes was tested together as generally the variability in species data (when constrained by environmental variables) is expressed by more than one canonical axis (Lep? & Smilauer PD-98059
2003). The forward selection procedure in CANOCO was used to determine which treatments added significantly to the explanation of the observed species variance. Relative contribution of grazing and soil to final species composition. To determine this, variance partitioning, a method to quantify the effects of two or more groups of ecologically distinct environmental variables (Lep? & Smilauer 2003), was used. The forward selection procedure in CANOCO was used to select a subset of soil properties that contributed to the explained variance (but were independent of each other) for inclusion in the ordination (Lep? & Smilauer 2003). These were soil pH, available magnesium, total nitrate and available phosphorus. Using a series of partially constrained ordinations on the floristic data (Borcard, Legendre & Drapeau 1992), the variation in these data was separated into four components; variation explained by (i) treatment alone, (ii) soil alone, (iii) these two factors combined and (iv) variation unexplained by either of these factors. All ordinations were tested with unrestricted Monte Carlo permutations (499 runs), with a Bonferroni-corrected significance level of 0.05/4?=?0.0125 (Borcard, Legendre & Drapeau 1992).