Showing posts with label Bayes. Show all posts
Showing posts with label Bayes. Show all posts

Monday, November 23, 2015

The Laws of Medicine - Siddhartha Mukherjee

Siddhartha Mukherjee is very smart, very thoughtful, and a very talented writer, with an ear for the telling anecdote.  (We have previously reviewed his Pulitzer Prize-winning The Emperor of All Maladies).  The Laws of Medicine is a focused reflection on how medicine has changed over the past 80 years, viewed through the lens of Mukherjee's desire to identify "laws", specific to medicine, that govern its practice.

The book is serious and thought-provoking but also rather charming....and it's a fast read.  It's definitely a library book, not a worthwhile purchase - it's almost pocket-sized, it's thin, and it has some lovely textless illustrated pages to add space for personal reflection between one chapter and the next.  Purchasing it ($16.99) would be a good way to support the TED Talks (from which this book was adapted), but I was happier to delight in Mukherjee's reflections, on the library's dime!  To offer a little substance to this review, here's a nice thought, snatched from the book:
Priors. Outliers. Biases.  That all three laws of medicine involve limits and constraints on human knowledge is instructive.
I definitely recommend this highly.

Thursday, July 17, 2014

Nate Silver's The Signal and the Noise

Gibson's novels deal extensively with the science of prediction, which he depicts as a kind of occult phenomenon or 6th sense (cf Laney with his uncanny nodal perception, or the Coolhunter Cayce). As usual, in exploring this theme, Gibson is masterfully extrapolating from current issues, such as the rise of and obsession with Big Data. My reading of this book, then, was driven by curiosity of the actual state of prediction today, especially as compared to its potential as envisioned by Gibson. Sounds like we have a ways to go! It was a very fast and enjoyable read, written in a fluent and conversational style. Recommend!

Dad's review here.

Sunday, May 18, 2014

The Signal and the Noise

This very fascinating book by Nate Silver, of 538.com and NY Times fame, was recommended by Jesse and a geneticist friend, Anil Menon - and it's a winner!  The book is all about predictions - in this era of "big data", one might think that we would be able to make more and more accurate predictions, yet Silver shows that this is not the case.  He looks closely at several fields in which predictions are crucial - weather, earthquakes, climate change and others - and shows that predictions have become more accurate in some of these (weather), but not others (climate change and earthquakes) - and discusses why this is so.  The book sounds very wonky - Bayesian statistics are discussed at some length - but it is exceptionally readable and enjoyable.   Silver's writing seems like speech - it probably was dictated - and he has a whimsical sense of humor that catches you by surprise - he's laugh out loud funny.  He discusses Isaiah Berlin and hedgehogs and foxes (see Berlin's essay) to describe effective predictors (foxes) versus ineffective predictors (hedgehogs).  Really absorbing and informative.  Very highly recommended.  I plan to re-read it soon!