Statistical Analysis of Biological Survival Data

Abstract

The objective of this study is to throw the light on applications of survival analysis in veterinary and biological sciences. In veterinary sciences especially in dairy farms, there are important factors in dairy herds which have a great role in their effects on another important factor which is called days open. There are many statistical methods used to model and analyse the data under these circumstances, but here some methods of survival analysis will be used because there are two types of data in the study (complete data and incomplete or censored data). The common statistical methods cannot be used to analyse the censored data. Some methods such as Kaplan- Meier method, Log-rank method and the Cox's proportional hazard model method will be used in this study. The data were obtained from different lactation records, covering the period between 2004 and 2007. These milk records are of U.S. Holstein cows belonging to Dina farms. The result showed that based on the K-M survivorship percentiles, overall median days open of dairy cattle was at (134 days). There were a non-significant difference between seasons, and a highly significant difference between years and lactation order. The result from Cox's proportional hazard model showed that lactation order increased the chance of pregnancy. Age at calving, days in milk, and season and year of calving decreased it. The result of testing Cox's model assumptions using a Schoenfeld residuals showed no correlation between partial residuals of the variables under study and rank of days open, so the proportional hazard assumption is satisfied.