A blind reanalysis of a random subset of NCI bioassay studies: agreement between rats and mice.

There are many problems in the statistical analysis of long-term rodent studies. This paper discusses the alternative approaches to these problems suggested by the previous three papers and compares the results of the different analyses on 25 rat and mouse studies. It is concluded that many problems remain unresolved and that alternative, seemingly reasonable decision rules can give rise to dramatically different results.