Scientific Reasoning, the Bayesian Approach

by Colin and Urbach Howson

Introducing the Bayesian account of scientific reasoning, this book looks at the role of statistics and probability in science. The authors contend that the classical system is flawed and demonstrate that the Bayesian approach avoids such difficulties and is not too subjective a method.

Genres: Philosophy, Science, Nonfiction, Mathematics

Paperback

First published October 1, 1988

Reviews

[Bastian Greshake Tzovaras · Rating 3 out of 5]

One notices that the first edition was published in 1989, as the authors go to great lengths to argue against frequentism and for their bayesian view. Which can be fun in some places and tiresome in others. But in general it makes a nice connection of everything bayesian and the philosophy of science.

recommended for: nerds who enjoy the philosophy of science and statistics.

[Nick Geiser · Rating 3 out of 5]

This is an interesting book and a compelling argument in favor of Bayesian probability as theory of inductive inference. The arguments against null-hypothesis significance testing especially persuasive. Unfortunately, there are so many typos and mathematical mistakes in this book as to border on the inexcusable. It's not clear whether this is the fault of the author or the publisher, but either way it is hard to take the book seriously.

[Foppe]

Review here.

[Hans Ostrom · Rating 5 out of 5]

A fascinating look at the inductive, educated-guess mode of reasoning. The Bayesian approach informed the work that Alan Turing and his co-workers did to break the German codes.

[Eric Lawton · Rating 1 out of 5]

I picked this up because I am very interested in both statistics and philosophy of science, so this should have been just what I was looking for.
It started off well, but half way through I decided it was not worth any more time because the book was far too rambling. "Stream of consciousness" is OK for James Joyce, but not for a book on a fairly complex pair of subjects. I'm fairly familiar with the notion of a normal distribution but OK, some people may need a diagram, if only for a quick reminder of a technical term they may have forgotten. But if anybody needs that, why, a few pages later, do we get a fairly complex formula for a bivariate density with no diagram but have to imagine "a more-or-less-pointed, more-or-less elongated hump over the x,y plane, whose contours are ellipses...". Similarly, if you don't know what a Goodman grue-alternative is, you won't find it in the index. I've read Goodman and know that it is "green today and blue tomorrow" but I'm still baffled as to how the rest of the text applies to the paradox, though the mention is consistent with similar name-dropping of several philosophers of science per paragraph in other chapters.
My charitable interpretation is that the book could have used a lot of editing but I wasn't about to do it on the fly. There are far better books on both topics.