3. MATERIALS AND METHODS
3.11 Statistical analysis
3.11.1 Summary of the variables considered in this study
The variables described in sections 3.7, 3.8, 3.9, and 3.10, and summarised in Table 3.6, were used in correlation analyses to determine relationships between performance measures, feeding behaviour, dominance, and activity.
Table 3.6: Summary of the variables used in correlation analyses with performance measures, grouped by type: feeding behaviour (all 1049 animals), dominance index (32 animals), and activity (32 animals).
Performance Residual feed intake
Average daily liveweight gain Liveweight Feeding behaviour Intake Meal frequency Feeding duration Intake rate Meal size Meal duration Dominance Index Majority criterion Binomial criterion Activity Lying Standing Walking Drinking Feeding
Using the salt block Lying ruminating Standing ruminating
3.11.2 Cohort correlations
Correlations between variables were developed using the data collected from the 219 animals in Cohort 2, and these methods were applied to the other four cohorts. Correlations using all animals in each cohort were calculated by regression analysis for each cohort using GenStat (Payne et al., 2009). Correlations were calculated for both ADG and RFI with each feeding behaviour characteristic (intake, meal frequency, feeding duration, feeding rate, meal size and meal duration) for each cohort.
3.11.3 Comparison of top 10% and bottom 10% animals
The performance and feeding behaviour of the 10% most (low-RFI) and 10% least efficient (high-RFI) individuals from each cohort were compared using analysis of variance (ANOVA) in GenStat. The cohorts were combined in a pooled within-cohort analysis (i.e. cohort included in the model), which allowed an overall comparison of the 10% most and 10% least efficient animals of all cohorts, presented in chapter 4. Comparisons using ANOVA were also made for the most and least efficient groups (10% of each) from individual cohorts, and these results are presented in Appendix II, Section 7.2.2. This analysis provided a numerical summary of
divergence and illustrated average values of RFI, ADG, LWT, intake and feeding behaviour for the most (low-RFI) and least (high-RFI) efficient groups.
In total, 208 animals were included in this analysis of extremes: 32 in Cohort 1 (16 low-RFI and 16 high-RFI), and 44 in each of the remaining cohorts (22 low-RFI and 22 high-RFI in each of cohorts 2-5). These animals were retained, mated, calved and were used for further research (e.g., postpartum liveweight change, methane production) as part of the on-going research.
3.11.4 Temporal feeding patterns
Analyses included the time of day that animals consumed their feed. The 24 h day was divided into eight 3-hour time periods (0000-0300, 0300-0600 h, etc.) and average feeding behaviour was summarised for each of these periods. Initial evaluation with Cohort 2 indicated differences between the 10% most and 10% least efficient groups, which prompted an evaluation of the 10% most and 10% least efficient animals in all cohorts using a pooled within- cohort analysis (as described in section 3.11.3). Cohorts were combined for simplicity and brevity and these results are presented in chapter 4, whilst the evaluations of each cohort separately are presented in Appendix II, Section 7.2.3.
A repeated measures analysis of data was conducted using AREPMEASURES procedure in GenStat, with time included as a factor, to determine whether there were differences in feeding behaviour between the most and least efficient animals (n = 208; 10% most and 10% least efficient animals in each cohort combined) during a 24 h day. All variables had significant interactions between time and efficiency group (p < 0.001), so this interaction was then explored at each 3-hour time period. The mean intake, meal frequency, and feeding duration of the 10% most efficient (low-RFI) and 10% least efficient (high-RFI) animals in each cohort were compared in each 3-hour time period using ANOVA in a pooled within-cohort analysis to determine feeding patterns over a 24 h day. The feeding patterns were also compared within each group (i.e., the change in feeding activity of each group over 24 h), hence two standard errors of the difference (SED) were calculated. Correlations of mean intake, meal frequency, and feeding duration in each of these periods were calculated against RFI and ADG for all five cohorts (n = 1049), but these are not presented as they showed similar patterns to those shown in the comparison of the most and least efficient animals (Section 4.6).
3.11.5 Dominance and activity
Dominance indices (DI) and activity profiles were calculated for a subset of animals (n = 32) in Cohort 5, as described in sections 3.9 and 3.10. The number of interactions per pen, and the zone and time of day (in 3-hour time periods) at which these interactions occurred were calculated over the 48 h of recording. As well as daily activity, proportions of time spent in each activity in 3-hour time periods were calculated. Correlations between variables of DI, activity, feeding behaviour, and performance were derived for each pen (GenStat), because the DI and activity of each animal was dependent on the presence and behaviour of its penmates. The slopes of the correlations for each pen were compared for each variable, and when there were no statistically significant differences (p > 0.05) between the regression slopes of individual pens, the slopes were pooled to give one correlation for all pens. These pooled slopes were used in this analysis where appropriate.
To determine whether there were differences in activity between animals of differing dominance status, animals were grouped into three dominance categories based on their DI: Low: DI < 0.4
Medium: DI 0.4-0.6 High: DI > 0.6
A repeated measures analysis of data was conducted using AREPMEASURES procedure and splines in GenStat, with time included as a factor, to determine whether there were differences in activity between the three dominance categories within pens (n = 8/pen) during a 24 h day.