mercoledì 1 novembre 2017

The art of lying with statistics and ecology

It is all too easy to say that statistics is the art of lying and that you can demonstrate everything by simply changing a parameter. You can actually change evrything bu cahnging a parameter, but that is true also if you change the method of staining of a preparate under a microscope. The problem is that biologists (an psychologists and physicians) think that statistics is a sort of artificial intelligence, not different from the supercomputer in The Hitchhiker's Guide to the Galaxy; where you enter the data and get the answer. Science is not that easy and if such a computer will ever be build it will anwser "42". Statistics is a tool, exactly like a microscope. If your put the wing of a fly under the microscope you will get a very different picture than if you put a leg - and that is exactly what you do when you change a parameter. The fact that most ecologists do not appreciate this is because they often cannot tell a wing from a leg: they download huge amounts of data from repositories and play with models without knowing much about  the data - they often put under the microscope the lef of a fly thinking it is a wing. 

Another problem is that statistics is expressed in formula, so that many mistake statistics for mathematics, which statistics is not. Mathematics is a language, statistics is again a tool. A language is a set of rules that allow to relate thing (often in very complex ways) and thus extends our midn, whereas a tool is something that extends our senses, like the microscope or the telescope. Fundamentally statistics is a tool that allows to "see" the randomness and noise in data, which is something that our brain has much difficult to appreciate. The simple regression model is for instance Y = a*X +error;; the difference between a statistic tool and a mathematical relation is just in the last tarm.

Nessun commento: