Handling Cause of Death in Equivocal Cases Using the EM Algorithm

摘要

The statistical analysis of animal bioassays for carcinogenicity often involves utilizing the cause of death of each animal. For example, the NCTR, IARC and ORNL all recommend collecting and using cause of death data. However, there is still considerable disagreement among veterinary pathologists as to the reliability of cause of death information. This has prompted recent recommendations for assigning cause of death in animal studies to allow for uncertainty on the part of the pathologist. The NCTR and other laboratories are now collecting data that contain acknowledged equivocal cases with respect to cause of death. The report describes a method for incorporating such equivocal information into the nonparametric estimation procedure currently used by the NCTR for analyzing chronic bioassays of carcinogenicity. A preliminary version of the report was presented at the national annual statistics meetings and a written version has been submitted for publication in Communications in Statistics.