How to Read Numbers: A Guide to Stats in the News (and Knowing When to Trust Them)

Description

'Even one glass of wine a day raises the risk of cancer’ ‘Hate crimes have doubled in five years’ ‘Fizzy drinks make teenagers violent’ Every day, most of us will read or watch something in the news that is based on statistics in some way. Sometimes it’ll be obvious - ‘X people develop cancer every year’ - and sometimes less obvious - ‘How smartphones destroyed a generation’. Statistics are an immensely powerful tool for understanding the world; the best tool we have. But in the wrong hands, they can be dangerous. This book will help you spot common mistakes and tricks that can mislead you into thinking that small numbers are big, or unimportant changes are important. It will show you how the numbers you read are made - you’ll learn about how surveys with small or biased samples can generate wrong answers, and why ice cream doesn’t cause drownings. We are surrounded by numbers and data, and it has never been more important to separate the good from the bad, the true from the false. HOW TO READ NUMBERS is a vital guide that will help you understand when and how to trust the numbers in the news - and, just as importantly, when not to. Dedication Title Page Contents Introduction Chapter 1: How Numbers Can Mislead Chapter 2: Anecdotal Evidence Chapter 3: Sample Sizes Chapter 4: Biased Samples Chapter 5: Statistical Significance Chapter 6: Effect Size Chapter 7: Confounders Chapter 8: Causality Chapter 9: Is That a Big Number? Chapter 10: Bayes’ Theorem Chapter 11: Absolute vs Relative Risk Chapter 12: Has What We’re Measuring Changed? Chapter 13: Rankings Chapter 14: Is It Representative of the Literature? Chapter 15: Demand for Novelty Chapter 16: Cherry-picking Chapter 17: Forecasting Chapter 18: Assumptions in Models Chapter 19: Texas Sharpshooter Fallacy Chapter 20: Survivorship Bias Chapter 21: Collider Bias Chapter 22: Goodhart’s Law Conclusion and Statistical Style Guide Acknowledgements Notes Also by Tom Chivers Copyright