The Art Of Statistics: Learning From Data by David Spiegelhalter (book review).
Statistics are nothing unless you understand how sampling and analysis is done and can draw real conclusions from them. With so many statistics bandied about on the news for covid-90, being able to understand the figures and how they are composed is a handy tool to have in your toolbag.
Statistics is essentially examining gathered information and seeing if they reveal a pattern. With ‘The Art Of Statistics: Learning From Data’, I do think author David Spiegelhalter expects you to have some basic knowledge of the subject although admits he won’t baffle you too much with the maths. Well, until you read the glossary.
Rather than go into the subject drily, he goes by examples that he’s worked on. As Spiegelhalter is chairman at the Statistical Laboratory at the University of Cambridge and a former President of the Royal Statistical Society, he has been involved in many cases that have had media exposure which means you will at least have heard of them. Seeing the proofs analysed does illustrate how the focus has to be made at the differences. Magnifying these areas also shows what you think is a slight difference is often much more. Think of that when you set your graph start points at zero that close to the mark where changes are more apparent at the top end.
I was wondering should I go through all of his examples but as I’m reviewing, it’s best for you to read them for yourselves. So I’ll do what I normally do and look at oddities. Even the people in the press wondering in the survey of how many sexual partners people had how men had vastly more than women when they should have been a lot closer. Even reading the results here, I’m surprised no one considered men might exaggerate their sexual prowess if only to not look like losers but there should have been cross-checking questions to cancel out exaggerations.
To be fair, the follow-up chapter says those who were filling in the questionnaire were linked to what they thought was a lie-detector so to give honest answers. Clearly, it couldn’t have worked or they wouldn’t have had the exaggerations but there are always people who think they can cheat even the real machines.
Oh, you shouldn’t be surprised that Gaussian curves I’ve often mentioned are covered although you will need to look under ‘normal distribution’. A lot of statistics do not require people answering questions but just adding up information as with a randomised controlled trail. Here you need to prove people react well with a new drug it needs to be balanced a blind test with those told the same thing but given a placebo just to show they aren’t just getting better by their bodies own idea of remission. When the medical research team themselves don’t know who is who, hence the blind testing, it reduces anyone manipulating the results and gets a clearer honest set of figures. I’m giving the details of that as any covid-19 inoculation or vaccination will depend on a similar test.
Where statistics come into their own is predicting future odds and testing against earlier examples for validity. I’m not so sure if using the Titanic as the best example for survival rates as the circumstances were a lot different to what we have today and lesson were learnt and applied. However, we are now in the realms of algorithms and borderline Artificial Intelligence so information can be tested faster and turned into graphics. I should point out that there are a lot of graphs in this book and make it easier to visualise than numbers for us but maybe not for computer programs. Such information is also used in gambling for predicting odds. I suspect that will be of more interest than the standard deviation and the error allowance that has to be made.
With marketing surveys, a lot depends on the questions asked. Although Spiegelhalter shows some of these, I do wish more time had been spent over them but suspect it might clue in potential test subjects of things they really need to be blind-sided about so you don’t give them information you think they want rather than honest replies. I’ve never been sure about graded questions of how much you like something because there’s a tendency to not know how to place things.
The ultimate question about any specialist subject books is do you come away with a better understanding of a subject after you’ve read it. With this book, I would give a definite yes. Statistics are an important part of our lives but I do think not many people know nearly enough about them as to their worth. Don’t shy away when you’re asked for them by the authorised statisticians as you’re becoming part of a representation that might help a larger part of the population and even help you with your life. It’s all in the numbers. Being a statistic is important.
GF Willmetts
October 2020
(pub: Pelican/Penguin/Random House, 2020. 426 page paperback. Price: £ 9.99 (UK). ISBN: 978-0-241-25876-7)
check out website: www.greenpenguin.co.uk
