How to Lie With Statistics (1954)
archive.orgAmazing that after 70 years a lot of these practices are still in use and people fall for them. My favorite example are bar charts with truncated axis that are made to show extraordinary growth. But, as is well known, data doesn't matter as much as how you present it (or how much you hide)
One of my favorite things published of all time! Helped convince me to become a Statistics major when going into college.
If anyone interested to listen to a podcast about the book https://nouswise.com/c/8a6439ca-fee7-4401-8ef3-b92a823cfd2a?...
This is probably the best book on statistics you can give to a student of statistics.
I was under the impression that it is now out of print.
My copy is from 1990 or so.
Don't even need it. Just have an AI tool summarize it, provide up-to-date examples, show where the arguments are overstated, list unaddressed areas for further development, and create flashcards and podcasts.
With a book like that, you don't even need to provide the source material.
I love this book. It has been so helpful throughout my career.
Lying with statistics is a subset of the practice of lying with numbers. My [least] favorite trick to lie with numbers is to provide numbers without any context (e.g. "this datacenter uses 1 million gallons of water per year").
Wouldn't necessarily call it a lie. More like implicature or maybe "Gricean expectation engineering".
A bit ironic, your comment could be viewed as using a similar technique. Given the ongoing controversy around water consumption in datacenters, choosing this particular example implicitly associates people raising concerns about datacenter water use with deceptive rhetorical practices, without actually making that argument explicitly or presenting a concrete context in which the supposed rhetorical device was used.
Of course, that's not necessarily your intention, only a point about implicatures.
This book is still in print, why are you sharing a pdf copy?
Why not? Darrell Huff died in 2001, and under any respectable copyright system his work would have ended up in the public domain in 1979, or at least with his death.
Corrected, thanks, though a 1954 copy, but a lot of copies online. A good one, nevertheless (i.e., something you might actually pay for a paper copy).
I like Computer Science because it's hard to lie with code.
Tell that to Claude.