Smith, Leonard A. (2003) Predictability past predictability present. In: ECMWF Seminar on Predictability, 2002, Reading, UK.
Predictability evolves. The relation between our models and reality is one of similarity, not isomorphism, and predictability takes form only within the context of our models. Thus predictability is a function of our understanding, our technology and our dedication to the task. The imperfection of our models implies that theoretical limits to predictability in the present may be surpassed; they need not limit predictability in the future. How then are we to exploit the probability forecasts extracted from our models, along with observations of the single realization corresponding to each forecast, to improve the structure and formulation of our models? Can we exploit observations as one agent of a natural selection and happily allow our understanding to evolve without any ultimate goal, giving up the common vision of slowly approaching the Perfect Model? This talk addresses these questions in a rather applied manner, and it adds a fourth: Might the mirage of a Perfect Model actually impede model improvement?
|Item Type:||Conference or Workshop Item (Paper)|
|Additional Information:||© 2003 LA Smith|
|Library of Congress subject classification:||H Social Sciences > H Social Sciences (General)
H Social Sciences > HM Sociology
|Sets:||Departments > Statistics
Research centres and groups > Centre for the Analysis of Time Series (CATS)
Actions (login required)
|Record administration - authorised staff only|