David Spiegelhalter

After watching Professor Spiegelhalter give a really engaging online lecture for the Royal Statistical Society (which he was president of in 2017-18), I thought it time to give his book a try. He writes very clearly and engagingly, so I definitely recommend it to anyone who is curious about statistics or has to take a course in it but doesn't feel very confident about maths. The only equations are in a glossary at the end; statistical concepts are explained in words with graphs and show more visualisations to assist. I used to teach introductory statistics (which had been rebranded as 'analytics') to students and would definitely have put this book on the reading list had it been available. I now work as a statistician, so the concepts were not new to me. In fact, I use most of them daily. For me the enjoyment was in the wide range of examples Spiegelhalter uses to illustrate his explanations.

Many such are concerned with health, including statistical analysis of Howard Shipman, a GP who became Britain's most prolific serial killer. The first chapter deftly shows how statistical tests could be used to demonstrate how unlikely it was that the number of death certificates Shipman signed was due to chance alone. It also notes that when the same test is applied across a dataset of all GPs, it immediately identified another one that was signing even more. The latter GP, by contrast, worked in an location with a high proportion of elderly people and was compassionately enabling them to remain at home until the end of their lives, rather than dying in hospital. A key lesson from the book is that data analysis requires contextual information to be meaningful and useful.

I particularly appreciated how well Spiegelhalter explained Bayesian methods, which I'm familiar with in theory but have never applied in practise. (I can only remember how methods work if I've actually used them for something.) My favourite chapter explained how Bayes theorem is employed in court cases and in estimating the likelihood that Richard III's skeleton had been unearthed. This chapter also covers how exit polls for UK General Elections are conducted, which is of particular interest less than two weeks after the last one. The 2024 exit poll correctly predicted a Labour landslide.

The last two chapters are titled 'how things go wrong' and 'how we can do statistics better'. Both are very sensible and measured. This quote elegantly encapsulates a sentiment that I periodically experience at work:

As Ronald Fisher famously put it, 'To consult the statistician after an experiment is finished is often merely to ask him to consult a post mortem examination. He can perhaps say what the experiment died of.'

Statistics is important and extremely useful. Take it from someone who did not do maths A-level and said at the time, "I'd rather stick pins in my eyes". Even if you're not keen on maths, you might end up fascinated by applied statistics in the social sciences like I did! Spiegelhalter is brilliant at communicating how they work and why they matter. show less

In all the COVID-19 books that appear on bookshelves, Covid by Numbers has to be one of the most unique. It’s about the statistics behind the COVID-19 numbers – not just rates of infection, hospitalisation and mortality but how numbers and modelling were used to determine lockdowns and other rules. It also looks at the unexpected effects of the COVID-19 pandemic, such as the lowering of influenza and car accidents in young people.

If you’re not a statistician, don’t worry. show more Spiegelhalter and Masters explain the statistical methods very succinctly and clearly. (Honestly, I have never seen logarithmic scales in graphs explained in such a simple way that makes sense). The emphasis is on what the statistics mean and how they were used to make decisions regarding masks, lockdowns and who to vaccinate first. If you understand words, the graphs and tables are explained clearly. If you prefer pictures, the graphs and tables are there too. The content is heavily focused on the UK (Australia doesn’t get a mention) although there is some comparison with EU countries and with the US. To me, this didn’t matter but it might if you’re looking for a more global summary. But by keeping the statistics close to home, the authors are able to provide links to what was happening at the time on the ground as one of those affected.

Of course any COVID-19 book is going to be dated as soon as ink is put to paper. Covid by Numbers covers 2020 and up to May 2021, so it’s pre-Omicron and Delta and pre-oral treatments for COVID-19. Again, this didn’t really matter to me because the authors went into sub-topics of COVID-19 I’ve always wondered about. What is the effect on the economy (spoiler: they didn’t predict rapid inflation)? What are some of the effects of the lockdown? Why did deaths go down below ‘normal’ levels at times? I found this all fascinating, as these questions hadn’t really been covered for me before. What this book also shows is how far we’ve come in the nearly three years since COVID-19 was first reported. We have vaccines and treatments and have moved on from lockdowns. People have seen more statistics than they did in the preceding decade and the SIR model of infection is well known, as are the meaning of reproduction numbers.

This is a good read for looking at the COVID-19 pandemic from a different point of view. I’d love to see an updated version in years to come.

http://samstillreading.wordpress.com show less

An excellent book. I wish it had been written when I was a student of Economics. Instead of just giving statistical formulae, it explains the background to using them. It brings it to life with copious real world examples, many of which the author had been involved. In particular, the enquiries into Harold Shipman and child heart deaths at Bristol Royal infirmary. He exposes the fallacies behind popular newspaper scare stories, such as cancer risks. He makes the subject matter amusing as show more well as fascinating, and never in an intimidating way. It's also interesting to know the characters behind the different methods, such as Fisher (confidence levels); Neyman and Pearson (type 1 and 2 errors); and Bayes(inverse probability). The author admits favouring the last. This book should be on the reading list of every student, journalist and politician. At a time of Covid, never has it been more important to understand the bases of the many statistics debated. show less

After being disappointed by a couple of other statistics books for general audiences (Stigler's Seven Pillars of Statistical Wisdom and Abelson's Statistics as Principled Argument), I finally found what I was looking for. The first few chapters were a bit on the simple side but certainly entertaining enough to keep my interest up. For me the real rewards were in the later parts of the book where the author discusses slightly more advanced topics such as P-values, confidence intervals and show more Bayesian inference. I had briefly acquainted myself with these topics in my university studies but did not really understand their meaning back then. Without going into their mathematics, the author provides a nice overview of what these statistical concepts mean and what they don't mean. This somewhat philosophical review of practical statistical methods should be very useful for anyone who wants to be able to judge statistical conclusions critically. In the final chapters of the book the author also discusses how scientists sometimes misuse statistical methods, how such misuse can be detected and how newspapers and their readers should understand new scientific results when they are expressed in terms of statistics. All of which is very interesting, so I recommend this book. show less