Science builds through an advance in knowledge. Knowledge accumulates when we compare a situation where something is present versus the same situation when that something is absent.

If someone hands you a cure for HIV which you then begin to inject into every HIV-infected patient you treat (forgetting everything you know about ethics and experimental design for the moment), and fifteen years later they are all still alive and healthy, you can be pretty sure that injection made some difference, can you not?

The reason for this is that, for many years, there has been a naturally-occurring control group: people with HIV and no known effective treatment.

Please, please, please never forget what I said about confusing tools with science. Tools are technology. Science is the creation of knowledge, and tools are used in this pursuit. Statistics and research design are tools. If you cannot make good decisions without them, you probably will not do much better with them.

Any time you have a "hunch" that you would like to test, always ask yourself what (from your store of professional knowledge) needs to be held equal in the control and experimental groups to insure that you get a good, clean test of your "hunch".

• Imagine the life of a member of the control group.
  • What else could happen to him or her that might produce the same result as your experimental treatment? How can you rule this possibility out?

  • Is it possible that he or she does not really belong in this experiment (e.g., false positive test for the condition of interest)? How can you rule this out?

• Imagine the life of a person in your experimental group.
  • What might happen in his life that would produce the same result as his or her not receiving the experimental treatment? How can you rule this possibility out?

  • Is it possible that he or she really does not belong in this experiment (e.g. false positive for the condition of interest)? How can you rule this out?