Want to learn the ideas in How To Lie With Statistics better than ever? Read the world’s #1 book summary of How To Lie With Statistics by Darrell Huff here.

Read a brief 1-Page Summary or watch video summaries curated by our expert team. Note: this book guide is not affiliated with or endorsed by the publisher or author, and we always encourage you to purchase and read the full book.

1-Page Summary of How To Lie With Statistics

Overview

Samples are only useful if they’re statistically significant. If not, then experimenters can manipulate their results until they’ve reached the desired results, which allows them to make sensational claims under the guise of scientific language.

Statistical thinking is critical for success in the modern world. As H. G. Wells once said, “Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write.”

Statistics are important to understand, but not many people do. This is because they don’t know how to think statistically. The book How to Lie with Statistics was written as a response to this problem in order to teach people how statistics work and why they’re useful.

Today, we are exposed to a lot of statistics. For example, juice is 26% more concentrated than before and four out of five doctors agree that toothpaste kills 23% more germs.

On the surface, these claims seem to be telling you about the products and services that you’re considering buying. But they are really just a ploy to get you to buy their product. The truth is more complicated than what they claim because there is no way of knowing if the statistics they provide are actually true or not.

In this passage, you’ll learn about the reasons why it’s safer to drive in bad weather than good weather. You’ll also find out why one magazine failed at predicting the 1936 presidential elections and how hard it is to count all of your beans.

Big Idea #1: It’s extremely hard to get a truly random sample.

Let’s imagine that we want to know how many beans are in Bob’s Bean Production Plant. The only way to know for sure is to count them all individually, but it would be a time-consuming and expensive process.

Fortunately, there’s a simple way to do this.

To make statistical estimates, you need to create a sample. A good sample is crucial because it’s the basis for drawing conclusions in statistics.

But for a sample to be representative, it must have two qualities: it must be large enough and random. We’ll talk about the size of samples later and focus on randomness first because only randomly selected samples can give you accurate information.

For example, if you’re interviewing 25-year-old women about how often they play guitar, you must select them randomly regardless of their income or social class.

It is easy to say that we should get a random sample, but it’s difficult to do. It takes time and money to count every bean by hand, and finding creative ways of reaching a truly randomized sample is also hard.

If you want to find a good sample of beans, it’s easy if they’re mixed up. You just pull out a handful and that’s your sample. However, what if the barrel wasn’t mixed? Then you’d only get white beans from the top.

If you were to base your sample on that and conclude that the barrel is full of white beans, you would have fallen victim to a type of bias known as “sample bias.” In the same way, non-randomized samples can negatively affect an experiment or study.

Big Idea #2: Bias can occur if the sample is not selected randomly.

Now that we know about biased samples, how can we avoid them? We can do this by using a strategy called stratified random sampling. First, divide the people into groups based on their commonness. Then pinpoint the proportion of each group in your population.

For example, if you’re trying to figure out the demographics of vegetarians, it’s difficult without a huge data bank at your disposal.

The second step is to get a random sample within each subgroup. For example, you could interview vegetarians of different ages or races. It’s important that the people in your sample are truly random, so it can be difficult to find groups like vegetarian under 18-year-olds.