....all models are wrong, so honest modelers report their uncertainty first and foremost....
- email signature used by Bill James
George E. P. Box famously said that “essentially, all models are wrong, but some are useful; the practical question is how wrong do they have to be to not be useful” (the first part is quite often quoted). My answer to his question is that honest modelers analyse and report their model uncertainty and error first and foremost, and honest models like PCSWMM provide ready tools for those analyses and reports.
I use the signature to reinforce the notion that computer models provide a drastic simplification of reality, and are in no sense accurate or true representations of real drainage systems. It is in this strict sense that they may be said to “wrong”. On the other hand, of course, PCSWMM models are more than useful – in many cases, essential. But they incur errors and modellers ought to analyse the uncertainties and errors; clients should be fully apprised of the likely implications. PCSWMM was written with this philosophy and these tasks in mind, hence the pithy signature in the title of this blog.
Background
Evidently Box wrote these words (or very similar) at least three times in at least two publications:
- In Empirical Model-Building and Response Surfaces (1987), co-authored with Norman R. Draper, p. 424, ISBN 0471810339, they wrote: “Essentially, all models are wrong, but some are useful.”
- In Box's paper, "Robustness in the Strategy of Scientific Model Building" (May 1979) in Robustness in Statistics: Proceedings of a Workshop (1979) edited by RL Launer and GN Wilkinson, a section heading on p. 2 reads: “ALL MODELS ARE WRONG BUT SOME ARE USEFUL”
- And again in Box and Draper, Empirical Model-Building, p. 74: “Remember that all models are wrong; the practical question is how wrong do they have to be to not be useful."
George Box
From Wikipedia, we learn that George Edward Pelham Box (b. 1919) was a distinguished statistician, who made important contributions in the areas of quality control, time-series analysis, design of experiments, and Bayesian inference. Box published the books Statistics for experimenters (1978), Time series analysis: Forecasting and control (1979, with Gwilym Jenkins) and Bayesian inference in statistical analysis (1973, with George C. Tiao). Today his name is associated with important results in statistics such as Box–Jenkins models, Box–Cox transformations, Box–Behnken designs, and others. He was born in Gravesend, Kent, England and trained as a chemist. During World War II, he performed for the British Army experiments exposing small animals to poison gas, and to analyze the results of his experiments, he taught himself statistics from available texts. He received a Ph.D. from the University of London in 1953, and received many academic and scientific awards.