Customer Reviews

How to Lie With Statistics

5 star:		(91)
4 star:		(26)
3 star:		(6)
2 star:		(3)
1 star:		(2)

 

 

This book, written in 1954, is just as pertinent today (perhaps even more so, as it's so easy to acquire statistics due to our current technology) -- Darrell Huff gives people the tools to talk back to statistics. Though there is a little bit about deliberate deception, in such things as "The Gee-Whiz Graph" (about how the graphical display of statistics can be twisted so that one can get any desired result, though the stats aren't changed), the meat of the book is regarding sound statistical reasoning, something that people today really need to consider.

For example, every person who listens to the latest survey showing a correlation between certain food and certain health problems or benefits should read "Post Hoc Rides Again", in which people erroneously leap from statistical correlation to a cause-and-effect relationship. An example given in the book is a report in which it was found that smokers had lower grades in college; ergo, said the researcher, smokers wishing to improve their grades should quit smoking! Of course, a statistical study showing that there's a "significant" relation between smoking and low grades doesn't show which causes the other -- perhaps educational failure draws people to smoke! My own theory would be that the =type= of person who is given to smoking is also given to not doing well in school; instead of cause and effect, one has a correlation from a shared, third (and unnamed) cause. One comes across these fallacies in the news =every=day=; I've been reading my online news, and in the science section I've already found two suspicious cause-and-effect reports. As Huff notes, it's not the statistics which are in question -- it's how they're used.

Some of the figures and examples used are funny due to their datedness (I love the picture of the surveyor asking a doctor what brand of cigarette he smokes, and the cigar-smoking baby just makes me smirk). It seems to me if you multiply every monetary amount by 10, you might get a better idea as to what it's worth (I don't know what it is actually worth, as I don't know what the inflation from 1954 is (another suspicious statistic)).

More to the point, with the help of this book, you need not have blind faith in the numbers or disgustedly throw all stats away. The mathematics of statistics guarantees them to have great power, as long as you know how to interpret them correctly. You might be pleasantly surprised to find that more common sense than math is involved in this book, but the truth is most modern abuse of numbers happens well after the numbers have been calculated. Of course, once you talk back to statistics people may think you're crazy; at least you won't be fleeced by false reasoning.

Statisticians hate the old adage "Lies, Damned Lies and Statistics", but statistical methods do have that reputation with the general public. There are many excellent accounts, some even understandable to laymen that explain the proper ways to analyze, study and report the analysis of statistical data. Huff's famous account is illustrative and well written. It gives the average guy a look at how statistics is commonly misused (either unintentionally or deliberately) in the popular media. Graphical abuses are particularly instructive. Readers should recognize that statistical methods are scientific and with proper education anyone should be able to recognize the good statisticians from the charletons. For now Huff's book is still a good starting place. As a statistician I hate the public image portrayed in the quote above. However, I do sometimes have fun with it myself. As I write this review I am in my office wearing a sweatshirt that reads "When all else fails manipulate the data."

A modern book by a consulting statistician on the same topic is "Common Errors in Statistics and How to Avoid Them" by Phil Good. If you enjoy this book take a look at Good's book also.

Although "How to Lie with Statistics" is a bit dated (having been written in the 1950's), the principles it puts forth are still valid today--if not moreso than ever--and the material is delivered in clear, concise, and even entertaining anecdotes and illustrations.

How often do you hear statistics bandied about in the media or used to try to prove some special-interest point? "Of course" the people quoting the figures must be right with numbers on their sides... until you look at just how those numbers were arrived at.

This book isn't truly a guide on how to lie with statistics, but it is an excellent text that informs the reader both how others will lie to them using statistics and on how to interpret the validity of purported statistical data.

My introduction to this book was by way of the 'required reading' list for my undergraduate statistics course. Bad first impression. But the book turned out to be fun to read, and enormously instructive. The class material for my college statistics course taught me HOW to do statistics, but this book gave me a good beginning understanding into the common methods of the abuse of statistics. Conversely, by implication, it also teaches how to present information in as truthful a manner as possible. The knowledge served me well as I further studied statistics at a graduate level, and continues to serve me as a Government Technical person, constantly working with statistical tools.

The book gives a good jump start into the interpretation of data presentations. Now, when I see a "Gee-Whiz Graph" I immediately know that the fluctuations shown in the line or bars are magnified, and I begin at once to look for the real difference (base 0) in the data points.

This book is living proof that learning can be fun. I highly recommend it to anyone working with or beseiged by data presented as graphs, averages, trends or any other such means. It will open your eyes.

-- with no equations. This book really is for every one. In fact, if you're a no-equations reader, this book will be especially helpful.

It shows all the little tricks that advertisers and propagandists, government agencies included, throw at you every day. One, p.85, is an impressive sounding news article about teachers' pay. At first, it looks as if a generous government outlay had doubled or tripled teachers' salaries. Looking closer, however, one sees an odd cluster of unrelated numbers flying in close formation. None of the numbers quoted has any bearing on any other, at least none that the article's reader can discover.

Duff also points out the fallacy of correlation. Oh, it's a useful enough measure, if (!) a number of mathematical requirements are met. It is not causation, however. For example, there is a strong correlation between a school child's height and the child's score on a given spelling test - taller kids do better. The fact is a lot less surprising when you see that first graders tend to be smaller than sixth graders, and tend to know fewer words. Maybe the example sounds silly, but no sillier than lots of the numbers in the news every day.

This is a quick and approachable read, and true even if the examples are now dated. Despite its name, this book really is aimed at honest people, readers who want real understanding of the data thrown at them, and presenters who want their numbers to be understood properly. And best, you don't have to be a mathematician to see what's going on.

//wiredweird

How to Lie with Statistics by Darrell Huff gives an explanation of common statistical errors. The book is clearly written and is understandable to a reader without a mathematics or statistics background. At only one hundred and forty two pages the book is a quick and easy read.

The book was originally published in 1954. The many copious examples were current at the time of writing, but are extremely dated now. Depending on the readers attitude this may be distracting, or faintly amusing. The advanced age of the examples does not make the text any harder to understand.

While the examples are dated, the concepts appear to be timeless. The same statistical manipulations still seem to be going on nearly fifty years later. The Author covers a wide range of statistical errors, or abuse. All of the types of errors will be familiar to anyone who pays attention to the news, or has seen an advertisement that uses numbers.

How to Lie with Statistics gives the reader the knowledge to detect common statistical skulduggery. If this knowledge were more widely spread, perhaps advertisers, political spinmiesters and sloppy journalists would not be able to get away with that sort of abuse.

For an excellent short introduction to the problems of polling, as well as other statistical nightmares, check out "How to Lie With Statistics" by Darrell Huff. This little book, which you can read in an afternoon, was written in the 50's and is *still* the definitive bible on how statistics can be misused.

It's fun to read, too, and I laughed out loud a number of times while reading it. A more accurate (but less catchy) title for the book would be "how other people lie with statistics, and how you can recognize it when they try to snow you." Each section describes a way that statistics or graphs are misused, and then gives real-life examples from advertisements or newspaper articles or political speeches of the author's day which illustrate the misuse in action. Sad to say, Huff's examples from the 50's look just like the crap we get shoved at us today. Some things never change.

The book only costs about $5, and from it you'll learn as much as an entire college course. Get a copy, read it, and lend it to friends. If I had to throw away all my books and could only keep a dozen, this would be one of the keepers.

"There are three kinds of lies: lies, damned lies, and statistics." This statement, made famous by Mark Twain, describes the persuasive power of numbers and how truthful statistics can be utilized to support inaccurate claims. This is exactly what author Darrell Huff shows readers in his book, How to Lie with Statistics. Huff explains how easy it is to manipulate statistics and what a common practice this is. How to Lie with Statistics informs readers so that they are not mislead, because "the crooks already know these tricks; honest men must learn them in self-defense."
Interestingly, Darrell Huff was not a statistician. He was merely an American writer, living from 1913 to 2001. He studied sociology and journalism at the University of Iowa. Duff was once editor of Better Homes and Gardens and Liberty magazines. During his career, he wrote hundreds of "How to" articles and books.
The first chapter entitled "The Sample with a Built-in Bias" explains the errors, intentional and unintentional, that can happen while producing statistical research results and how these errors lead to influenced or inaccurate conclusions. Huff says that much of the information we read in newspapers and magazines is incorrect because the sample used is not a representative sample--one in which every source of bias has been eliminated. To support his theory, he uses an example in which Time magazine claimed that "the average Yaleman, Class of '24 makes $25,111 a year."
Huff informs his readers that there are three different types of averages and if one type is used incorrectly it can cause misinterpretation of the data and inaccurate conclusions. The word average is very loose term, not giving a reader any real information unless it is specified as to what kind of average it is--mean, median, or mode. The mean is the mathematical average or the sum of all the numbers in a series divided by how many numbers are in the series. The median tells that half the numbers in the series are above and half are below than that number. The mode is the most common number within the series. Huff provides the example of the company that boasts that their average annual pay is $5,700. This number is actually the mean. The median is only $3,000, meaning that only half the people at the company are making more than this. Furthermore, most people at the company are making $2,000, which is therefore the mode.
Huff shows how graphs are often modified to deceive readers. Misrepresentation arises when the visual representation of the data is contradictory to the numerical representation. This often happens when a person wants to "'win an argument, shock a reader, move him into action, [or] sell him something." Huff uses the example of how national income increased ten per cent in a year to demonstrate how to take a graph and present the conclusion the author desires. In both pictures below, the numbers and curve are the same. The only difference is that the bottom of the graph has been cut off, giving the impression that the line has climbed halfway up the graph.
Huff teaches his readers how to defend themselves from deceit by learning how to talk back to a statistic. The first thing he says to do is ask, "Who says so?" Often the name that is cited is not supporting or authenticating the information. It is also important to look for both conscious and unconscious bias. The next thing to do is look for how the person reporting got their information. Data, and therefore statistics, will be very inaccurate if it is drawn from a biased sample, such as those that are too small, not representative of the whole group, or that is self-selecting. The reader must search for what's missing. The absence of relevant information, such as the type of average, the base of an index, or the factor that is causing a change to occur, are clear indicators that the information may not be completely truthful. The reader needs to make sure the author has not switched subjects between the raw figure and the conclusion. Finally, the reader needs to ask himself, "Does this make sense?" If it contradicts common sense, such as an impressively precise figure, then it probably it not true.
One misperception that Huff tries to educate his readers on is the relationship between two things or the lack there of. The author says that if there is a correlation between any two incidents, meaning that a relationship is present between them without providing insight into the direction of the relationship, it does not mean that one causes the other. For example, event one could be caused by event two, event two could be caused by event one, or event one and two could be caused by an unknown factor, event three. Huff encourages readers to question relations between things or events.
How to Lie with Statistics is an outstanding book complete with tons of examples and amusing pictures. Although written over fifty years ago, this book is not outdated. Because the book is easy to read and understand, it's no wonder that it is the best selling statistics book in history. How to Lie with Statistics provides readers with a new perspective and point of view when looking at information.

This charming, cynical, wittly little book exposes many of the cheap tricks, bogus techniques, and cunning deceptions employed by Madison Avenue, politicians, corporate PR departments and even -- dare we admit it? -- scientists. I am surprised to find that some statisticians here have panned this book. Do they imagine their craft is only used for the good of mankind? That their colleagues are all upstanding citizens who would not think of deceiving members of the public, employers, stockholders, or Members of Congress? Yes, the techniques described here are simple, and any professional should see through them. For that matter, any banker should see through common stock swindles and Ponzi schemes, yet in the business section you will often read about prestigious banks and long-established investment firms that were taken in by these schemes and robbed of huge sums of money. Experts fall for stupid tricks too. Anyway, the book was written for the general public, not for professionals. And the professionals should remember Huff's main point:

"The fact is that, despite its mathematical base, statistics is as much an art as it is a science. A great many manipulations and even distortions are possible within the bounds of propriety. Often the statistician must choose among methods, a subjective process, and find the one that he will use to represent the facts. In commercial practice he is about as unlikely to select an unfavorable method as a copywriter is to call his sponsor's product flimsy and cheap when he might as well say light and economical."

My mother was a professional statistician. She recommended this book to me at a tender age. I am not much good at mathematics but I never forgot the definitions of average, median, bias, error rate, causality and correlation. The book shows several ways to plot a deceptive graph. It shows how to select a biased sample, and how to take meaningless, small, random fluctuations and blow them up to look like a trend. This may seem a little subversive until you remember, as the author puts it, "the crooks already know these tricks; honest men must learn them in self-defense."

The illustrations by Irving Geis are cute.

The book is in its 46th printing. Its staying power can be compared to that of the other great short manual, "The Elements of Style."

For someone who has not read this book, the disparity in reviews is probably striking. Those literally looking for a statistics course, or especially statisticians, will be disappointed. This book is about understanding and interpreting statistics from a practical standpoint. Ignore the statisticians - they are being sensitive about how their work is abused. Ignore the students looking for a statistics book - that isn't the intent of Huff's work.

This is one of my favorite books of all time. It helped me become a much more discerning consumer of news. This book is intended to help you better comprehend stat's that are thrown at you to make a point - which essentially means all stat's. Unfortunately, many times this is done is disingenuous ways, and this book will help you see through such foolishness.

In short - buy this little book. It is one of the best book values you will ever find.