Influence of sample size and experimenter on reliability of measures of avoidance distance in dairy cows
In the present study we evaluated the influence of different sample sizes and different experimenters on the reliability of measures of avoidance distance (AD) at farm level. On 29 dairy farms the AD of 55–100% of the cows was assessed by two different experimenters (E1 and E2).
For both experimenters the herd median of AD (ADME) and the percentage of animals that could be touched (Touch%) were calculated. The reliability between experimenters was assessed by Spearman rank correlation coefficients. To assess the influence of sample size on reliability of AD, the tested
animals were randomly divided into two halves (H1 and H2), and ADME and Touch% were calculated for both halves and correlated with each other, with total ADME and total Touch%, and with the behaviour of the milkers. All measures of AD were highly correlated between experimenters (ADME rs
= 0.86; Touch% rs = 0.81). On farms with a higher value for ADME, however, some discrepancy was found between experimenters in ADME and Touch%. Smaller sample size reduced the number of significant correlations with milkers' behaviour. AD of H1 and H2 correlated only moderately
(rs = 0.38–0.43). In sum, smaller sample size reduced reliability and validity. Between-observer reliability of AD was relatively high, but there may be some observer influence. Further investigations are necessary to optimise the measures.