Table 1.
Tail injury scoring categories.
Fig 1.
Mean proportion of pigs from the 15 outbreak groups with different tail injury scores (tail damage) in the weeks before and after an outbreak.
0 No damage, 1 Flattened, 2 Red, 3 Bite marks or scratches, 4 Wound.
Fig 2.
Bar graphs of 3D data validation by a human observer.
926 automatic 3D tail detections were checked by eye and described as curled, high loose, low loose or tucked (‘short tails up’ occurred only 16 times so data are not shown). Data are grouped by the visual observation categories, using different colours for the 3D tail categories. Note that 0° = 3D 0 (low tails), > 0° to 30° = 3D 1 (part-raised tails), > 30°to 60° = 3D 2 (raised tails) and > 60° to 90° = 3D 3 (high tails). Numbers on the bars show the frequencies.
Fig 3.
Proportion of 3D tail detections of low tails (3D 0) on the days before and after an outbreak.
Data are shown for the 15 outbreak groups, and each line indicates a different group (designated Alpha to Victor).
Table 2.
Relationship between tail damage scores and 3D low tails data.
Data shown are the coefficient of effect (± standard error) for a series of Linear Mixed Models using the proportion of pigs with the various tail injury scores as the response, and proportion of 3D 0 low tails as the predictor (fitting batch/group/date as the random effect model). Data for all groups and all days were used in these models.
Fig 4.
Proportion of 3D low tail detections (3D 0) on the days leading up to an outbreak.
Data are shown for the 10 outbreak groups for which there were at least 2 days of data between days -1 and day-7 (week -1) and also 2 days of data between days -8 and -14 (week– 2). Each line shows data for a different group.
Fig 5.
Estimated means (± s.e.) for the proportion of 3D low tail detections (3D 0) during the 2 weeks before and the 2 weeks after a tail biting outbreak.
Week -2 was days -14 to -8, week -1 was days -7 to -1, week +1 was days 0 to 6 and week +2 was days 7 to 13. Data were from one outbreak and one control (non-outbreak) group from each of the 8 batches. For control groups, the weeks were assigned by using the outbreak day of their contemporary outbreak group from the same batch. Estimated means were generated from a Linear Mixed Model of Growth Stage (weaner or grower) + Week + Outbreak Vs Control + Week*Outbreak Vs Control, with Batch/Group/Day as the random effects. The * indicates a significant (p<0.05) difference between Outbreak (indicated by red squares) and Control groups (blue diamonds) at the indicated time point (based on Least Significant Difference testing).