The scientific design of experiments has a long history in statistics, engineering, chemistry, physics, biology, medicine, and most scientific areas. There are common issues across disciplines. This talk will describe how the Bayesian framework is particularly helpful in considering design problems across disciplines. It allows for the use of external information through a prior distribution and a loss function. The design of medical experiments with human subjects is made particularly difficult because of ethical issues, as well as scientific issues. The loss function can represent ethical constraints and the prior distribution can be used to quantify ethical issues, such as the presence or absence of equipoise which is an ethical requirement for randomizing subjects to treatments.
Several problems in the design of experiments will be reviewed including real examples from animal bioassay studies, reliability testing, randomized clinical trials and early phase trials for drug development with human subjects. Some novel ideas and methods will be described.
This talk will be aimed at a general undergraduate audience with very little background in statistics. Finally, an overview of the role of mathematical and statistical scientists in medical research will be briefly provided.