Stayability of Dairy Cattle: Models with Censoring and Covariates

This paper discusses implications of using the exponential distribution to analyze dairy cattle survival data when the data set includes only actual survival times and when it also includes censored survival ages. A disturbing inconsistency in a published analysis of censored data is a direct consequence of a wrongly assumed exponential distribution for the data. The data are reanalyzed by Weibu]l and lognormal distributions, two other widely used survival distributions. Methods for choosing an appropriate distribution are discussed.