Screening for disease and dishonesty
This is a rather late announcement of pages we have put up on the use of screening tests. Using lie detectors, breast cancer and HIV screening as examples, we show how an apparently accurate test, when applied to a group of people in which only a small proportion have the thing you are trying to detect, will generate many false positives.
This is a standard problem: the technical name is Bayes theorem but it appears in many forms. Future examples will include the use of bood tests to try and catch doping athletes, and risk assessments to try and identify people on probation who are likely to commit serious crimes.