Nerd Corner: “The Art of Statistics, Learning from Data”

Most people outside the UK may not be aware of who Harold Shipman was. He was a British GP who was convicted of murdering 15 of his patients though it is widely believed that he murdered over 200 patients before he was finally caught. Prof Spiegelhalter gave evidence at the Shipman Inquiry and it was this example that he started the book with, maybe slightly macabre but the graph that he used to present the salient features of the case was simple yet so effective. It definitely captured my interest and set the tone for what I could expect from the rest of the book.

Prof Sir David Spiegelhalter is the chair of the Winton Centre for Risk and Evidence Communication and the Winton Professor for the Public Understanding of Risk in the Statistical Laboratory at the University of Cambridge. This is the second edition of the book which was released about a year ago in 2019. It is almost devoid of the complicated mathematics that one may expect from an eminent statistician like Sir David while still elucidating the concepts that lie at the heart of the discipline.

The book covers a range of topics that is arranged in a very interesting way. It is not a book that focuses on a particular topic like “Algorithms, Analytics & Prediction” which could be a book in itself. Rather it is a book that tackles many of the topics that modern day statisticians are likely to encounter in the course of their work and not just “The Bayesian Way” and “Samples and Populations” but also “How Things Go Wrong?” or “How We Can Do Statistics Better?”. It is a supremely practical book with technical robustness. I can assure you that no matter where you stand in the spectrum of statistical expertise, there will be something that you will take home that you didn’t know earlier.

Not only is he is an renowned statistician but he is also an outstanding communicator. After all he is the chair of the Winton Centre for Risk and Evidence Communication. No book that is dealing with a subject such as statistics can be completely devoid of technical language but it does so in an extremely lucid fashion. What makes the book so unambiguous is his liberal use of examples and thereby cutting out the need to talk about concepts in an abstract way. Prof. Spiegelhalter is best known for his contributions to medical statistics and especially for championing the use of the Bayesian Approach and does use a number of examples from his work in this area but done so in a way that is easily understood by all. Having worked in data science for over fifteen years, I do know that I myself have struggled with communicating certain concepts to my colleagues and clients and I think this book has really shown us what good communication of statistical concepts should look like.

All in all, I feel that this is a book that all who work in this trade and employ statistics in any form will benefit from reading. In fact I would say that maybe this would a good book for someone who is seriously considering mastering statistics to get a hold of. He ends the book with some practical advice from “senior statisticians who, mirroring this book, are keen to emphasise the non-technical issues that are generally not taught in statistics courses.” Of the ten points that Sir David and his colleagues focussed, I have picked what I consider the most important. “ Statistical methods should enable data to answer scientific questions: Ask ‘why am I doing this?’, rather than focusing on which particular technique to use.”