Statistical analysis

Two statistical procedures were employed in the analysis. The effect of employing different FU upon the estimated environmental efficiency of dairy production systems was assessed using ANOVA employing the following GLM, y _ij=µ+S _i+Y _j+ε _ij where y _ij was the total GWP of the dairy production system per FU (LSU, ECM, MS, Land_farm, Land_total, ECM/Land_total), µ the overall mean, S _i the fixed effect of dairy production system (LFC, LFS, HFC, HFS), Y _j the random effect of calendar year (2004–10) and ε _ij was the random error term. Fisher’s tests were used to assess the level of significance of contributing effects, and differences between dairy production systems were determined by conducting pairwise comparisons using the Tukey’s method.

The ability of different FU to classify different production systems’ relative environmental efficiency was assessed using rank analysis. Using year as a repeated measure, systems were assigned a rank value from 1 to 4 in order of their relative emissions intensity, with rank 1 having the lowest GWP per FU and thus most efficient system. Rank analysis was performed using the Kruskal–Wallis’s test, a non-parametric equivalent to ANOVA. Significant differences between any two systems were assessed using the Mann–Whitney–Wilcoxon’s rank-sum test. The rank analysis was repeated when referencing the GWP of systems to each of the six FU employed in this study. All statistical analyses were conducted using Minitab 16.