2.7 | Statistical analysis
The effects of three feeding patterns on muscle fatty acid composition were compared by one-way analysis of variance (ANOVA) followed by Duncan’s multiple range test at the 95% confidence level. A rarefaction curve and Venn diagram based on OTU matrices were generated using Mothur 1.30.2. Principal co-ordinates analysis (PCoA) and distance heatmap plots were employed to visualize variation in botanical community across samples. To examine trends in RRA of items in the sheep diet, the data set was reduced to RRAs of the twenty most dominant taxa, which together comprised over 97% of all plant taxa sequences. Then, we conducted two-way ANOVA using GLM with Duncan’s multiple range test to identify significant differences in RRA of predominant plant items between grazing patterns and grazing periods. These statistical analyses were performed with SPSS Statistics version 17.0 (IBM, Armonk, NY). The fatty acid and RRA profiles were reported as mean values ± SEM.
Additionally, multivariate statistical analyses were conducted to interpret the relationships between variables and treatments using CANOCO for Windows version 4.56 (Leps, & Smilauer, 2003). The main gradients of variation in grazing treatments were described by principal component analysis (PCA). Redundancy analysis (RDA) was performed to determine the responses of herbage taxa and fatty acids to grazing treatment. The treatment combinations of herbage taxa (based on RRA metrics) and the treatment combinations of muscle PUFAs at the end of growing season were used as response variables for PCA and RDA. Monte Carlo tests were conducted in the RDA with restricted random permutations of samples. The correlation matrix was calculated in both PCA and RDA. Pearson correlation coefficients were also employed to analyze the relationship between diet taxa and muscle PUFAs. P < 0.05 was considered statistically significant. Figures were generated with OriginPro version 9.5.1 software (OriginLab Corporation, USA).