Although the title is ostensibly sinister, Darrell Huff’s “How to Lie with Statistics” is anything but. In medicine, we are faced with complicated statistics and “statisticulators” on a daily basis. And as the field of data science and statistics grows, so too does the complexity of these “statisticulations”. A statisticulation, defined by Huff, is “misinforming people with the use of statistical material” and, unfortunately, this is becoming all too common in the profit-driven world of medicine. With carefully crafted “non-inferiority” trials and overpowered industry-funded superiority trials cropping up in the literature, it would easy to give up on statistics altogether; but it’s imperative that we don’t. The key is harnessing the ability to identify the subtleties that statisticians use to misguide. As Huff eloquently states in his book, “The crooks already know these tricks; honest men [and women] must learn them in self-defense.”
How to Lie with Statistics was written for an audience with little to no background in statistics. There are no mind-numbing equations and, in the few instances where Huff needs to explain theories behind various statistics, he does so in a way that utilizes real-world examples that are easily comprehensible.
Each chapter was written so that it focuses on one individual statistical concept (e.g. bias, data visualizations, post-hoc fallacies, etc). Although the chapters were written as silos that can be read alone, together they do weave a coherent foundation of basic statistics. The final two chapters summarize everything into an easy to use tool to critically question statistical information we meet in our everyday life or, as Huff states,
“how to look a phony statistic in the eye and face it down; and no less important, how to recognize sound and usable data in that wilderness of fraud”.
Huff advocates in arming ourselves with knowledge to disarm false data by asking five critical questions when faced with any new statistical data.
- Who says so?
- Does this person have anything to gain from these conclusions?
- How do they know?
- Is the sample biased? Is the sample large enough to allow any reliable conclusion? Was the test performed appropriate to answer the question being asked?
- What’s missing?
- Does the paper tell the whole story? Do they leave out important information?
- Did somebody change the subject?
- Do the conclusions reported agree with the data presented?
- Does it make sense?
- Consider the conclusion. No, really consider it. Does it make logical sense? (Bayes Theorem)
Medical Education Relevance
The purpose of this book is to expose the reader to some of the subtle techniques data can be reported to make dull conclusions glamorous or, worse yet, transform negative conclusions into positive ones. This makes the book very useful as an adjunct to any existing introductory statistics curriculum. After all, medical practitioners do not need to become skilled statisticians, but must have enough background and understanding of statistics to apply clinically. Additionally, educators in medicine are not tasked with turning learners into mini-statisticians, but rather with turning them into knowledgeable consumers of evidenced-based medicine.
At just 144 pages, the book is short enough to be read in an afternoon – perfect for the busy and demanding schedule for medical trainees! This is great for those learners who especially have developed a distaste for the subject of statistics. Without becoming too burdensome, the book can create an important background and foundation in basic statistical knowledge.
This book was recently brought back into the limelight last month after the US Department of Veterans Affairs (VA) announced that they have banned its use from all future trainings. Was this ban due to dated information or bad statistical theory? No. “Significance” magazine reported that Tim Huelskamp, a VA House Committee representative, argued that the title of the book may create misperceptions to the lay public. 1 “Anything that creates a negative perception of the VA in the minds of veterans makes fulfilling our mission more difficult.” He later went on to say “I have not read and am not commenting on the merits of this book.”
Below is a list of additional stats-focused titles for the casual reader:
- Wheelan C. Naked Statistics: Stripping the Dread from the Data. W. W. Norton; 2014.
- The author of this book calls this “a homage to How to Lie with Statistics”. This is a fantastic adjunct to Huff’s work; it goes deeper into sampling errors and has a fantastic chapter on common mistakes made with regression analyses. Highly recommended.
- Goldacre B. Bad Science, Quacks, Hacks, and Big Pharma Flacks. Faber & Faber; 2010.
- Goldacre B. I Think You’ll Find it’s a Bit More Complicated Than That. Fourth Estate; 2014.
- Newman DH. Hippocrates’ Shadow. Simon and Schuster; 2009.
Here is a PDF of the book available for online.
* Disclaimer: We have no affiliations financial or otherwise with the authors, the books, or Amazon