Assessing the accuracy of modelling weight gain of cattle using feed efficiency data

Different methods of estimating weight gain were compared for accuracy and utility, using the amount of error variation from fitting the residual feed intake (RFI) model. Data were collected on 1481 cattle of temperate and tropically adapted breeds, feedlot-finished for the domestic (liveweight 400 kg), Korean (520 kg), or Japanese (steers only; 600 kg) markets. Cattle were tested in 36 groups over 4 years. The aim was to estimate weight gain over the period feed intake was measured, which was at least 49 days and averaged 63 days, including time for animals to adapt to the automatic feeding system. The different estimates were derived from linear and quadratic regressions of weight over time fitted to: F1) all weighings in the feedlot and F2) all weighings in the feedlot excluding atypical records in the first few weeks following feedlot entry. More complex linear and quadratic models were also fitted to weighings when feed intake was being measured, using the amount of feed eaten on the day of weighing, and previous days, to adjust for gut fill. Finally, a random regression model including general trends in the growth of each animal and short term measurement error was fitted to dataset F2 to estimate weight gain for the period feed intake was measured.

The RFI equation: feed intake=intercept(s)+β_w*mean(weight^0.73)+β_g*weight gain+error (i.e. RFI) was fitted using the different weight gain estimates. Based on mean squared errors from fitting this equation, longer measurement periods generally resulted in more accurate estimates of weight gain. The increased accuracy from using all weight measurements in the feedlot outweighed the loss from not measuring over the desired time interval—i.e. the period for which feed intake was measured.