Error is unavoidable in prediction. And it is quite common, often sizable, and usually consequential. In a clinical context, especially when dealing with a terminal illness, error in prediction of residual life means that patients and families are misinformed about their illness, that they may take foolish actions as a result, and that they may be given inappropriate or needlesly painful treatments or denied appropriate ones. In meteorology, error in prediction of storm paths or extreme events can have devastating consequences. In finance and economics, major policy decisions are taken on the basis of predictions and forecasts. Rational approaches to reduce and assess error in prediction are presented. Ideas are introduced how to relate these statistical strategies with clinical and medical concepts in particular and how to integrate ideas from apparently different areas.
Cite this article
Leonhard Held, Robin Henderson, Ulrich Mansmann, Statistical Issues in Prediction: what can be learned for individualized predictive medicine?. Oberwolfach Rep. 7 (2010), no. 1, pp. 217–251DOI 10.4171/OWR/2010/06