Effect of missing data on the estimate of average daily feed intake in beef cattle

Abstract

The effect of missing feed intake records on the estimate of feed intake in cattle over a 70-day test period was studied. Two datasets, containing several cattle breeds at different ages, were analysed. The first dataset comprised 300-day-old Angus cattle, while the second dataset consisted of 900-day-old Limousin and Jersey crossbred cattle. A first-, second- and third-order polynomial and a non-linear function were fitted to the intake data. The 3 polynomial functions showed minimal differences in their goodness of fit to the data, with the non-linear function explaining the least variation in intake. Both datasets were analysed over a 70-day test period and shortened tests were compared to the full 70-day test. Tests were shortened by randomly deleting intake records or periodically deleting data on a daily or weekly basis. The effect of missing data on feed intake estimates was evaluated using Pearsons correlation and t-tests. The results suggest that a greater proportion of data can be removed randomly and daily, compared with missing data in weekly blocks. However, when data are removed in weekly blocks, removal in the middle of the test period gives a more accurate estimate of feed intake. It appears that in young growing animals, estimation of daily feed intake is more sensitive to missing feed intake records than for older cattle.