How Not To Be Wrong

By Jordan Ellenberg

Kind of in the same vein as "The Signal and the Noise" Not that it's a clone. More like an intellectual cousin. How Not To Be Wrong discusses how math has been used as a tool to guide us through life, the limitations of what we can know, and how we can get a feel for the size and shape of the unknown, and how to take action when we are unsure.

"For this is action, this not being sure!"

"Error will bring us to the truth more quickly than vagueness."

There is a great chapter on smoking-- does smoking cause lung cancer, or does lunch cancer cause people to smoke. The answer is obvious now. Not so in the 1950's. While the statistics said that smoking was correlated to cancer, the mechanism was not known. Since correlation is not causation… since 90% of all statistics are made up… how do you prove that smoking causes cancer? How much data do you need? And what type? How do you show that you are not biased in your experiments? How do you make progress?

Several times the author touches on this theme. A mathematic process can give you a result if you give it any set of numbers. But, does that process make sense? Are you asking a question the data can answer?

Sometimes the answer is "I don't know" Randomness and noise can have very subtle effects that our brain erroneously thinks are patterns.

In this situation, can you measure the size and shape of the unknown? Control it's distribution and variability? This is preferred to deluding yourself into believing that you can create an answer if you torture your data enough.