ABSTRACT
As evidenced in the previous chapters, there are a great many statistical approaches to the analysis of drug safety data. Not all methods are good for all problems and as we have seen, methods which exploit dynamic exposures within individuals may well provide the greatest promise for future work in this area. Nevertheless, there are some approaches, often widely used in practice, which should be avoided as they are so prone to bias, misleading, and uninterpretable results that they are better off left alone. In this chapter, we try to highlight what these methods are and why they should be avoided and for what purposes.
