Anyone who has known me for more than an hour has probably heard me recite my favourite quote:
“Essentially, all models are wrong, some models are useful.”
Those wise words come from the statistician (and philosopher of science) George E. P. Box, and they may seem an odd thing for me to repeat so often, given that I’m the CTO of BOXARR, a company that makes software for building boxes-and-arrows models.
This week I had occasion to re-read Dr. Box’s paper from whence the original ideas in this quote come (to include a discussion of it in my upcoming book), and no one says it better than Box himself, in the abstract of that paper:
“Aspects of the scientific method are discussed: in particular, its representation as a motivated iteration in which, in succession, practice confronts theory, and theory, practice. Rapid progress requires sufficient flexibility to profit from such confrontations, and the ability to devise parsimonious but effective models, to worry selectively about model inadequacies and to employ mathematics skilfully but appropriately.”
All models are wrong, at least when they are of something that is worth modelling (e.g., something of sufficient complexity). In my mind, that’s precisely why we created BOXARR.
I think the ability to easily and flexibly make useful models, then “worry selectively about model inadequacies” and “employ mathematics skillfully but appropriately” is what BOXARR is all about.
Box’s paper is a really worthwhile read, as is a related paper where he refines my favourite quotation into the distilled form I so often repeat and discusses the statistical mathematics of why all models are wrong, and some useful, in detail.