Estimating meal criteria for meal pattern analysis of dairy cows

Extract

Meal pattern analyses depend crucially on appropriate estimates of bout or meal criteria (i.e., the longest non-feeding intervals accepted as part of a meal). Bout criteria are frequently estimated after fitting a ‘broken-stick’ to the un-transformed, the log-transformed, or the log-transformed cumulative (log-survivorship), frequency distribution of between-feeding interval length. We know of no biological justification for fitting a broken stick to the frequency distribution of short intervals between feeding events and can, therefore, not interpret ‘criteria’ obtained that way. The methods that fit a broken-stick to log-survivorship or log-frequency curves are based on the implicit assumption that the probability of an animal initiating a bout is independent of the time since the last bout. Only then will the length of intervals between bouts be distributed as a negative exponential that appears as a straight line after log-transformation. However, the satiety concept predicts that this probability is not constant but increases with time since the last meal.