Forgot your password?
typodupeerror
User Journal

Daniel Dvorkin's Journal: Lies, damned lies, and ... oh no, you're going there. 1

Journal by Daniel Dvorkin

[cranky rant warning]

"Lies, damned lies, and statistics." It's coming up again with depressing frequency, being used as an argument instead of a snide observation.

Okay, here's the thing. Can you lie with statistics? Sure. Statistics is a branch of mathematics*, and math is a language; you can lie in that language as easily as in any other. Does this mean all statistics are lies? No more than all statements in any language are lies--and if you believe that, you've gone so far down the rabbit hole of anti-intellectual mysticism that you'll probably never find your way out.

Meanwhile, in the real world, and in the ever-expanding torrent of data we have about that world, statistics as a discipline is pretty much the only hope we have of understanding anything. The low-hanging fruit has been picked. The equations we learn in Physics 101 are as valid as they ever were, but they're not nearly enough. No matter how certain you think you are, no matter how many times you repeat your experiment and get the same result, if you don't do the statistical tests you don't actually know whatever it is you think you know. And if you do the tests--well, you may still be wrong, but you can at least quantify your uncertainty. And you have to do that, because you can always be wrong.

None of this is meant to defend the misuse of statistics, any more than as a writer I'd defend the misuse of natural language. People can and do wilfully misinterpret statistics, or cherry-pick them, or just outright make them up, and those are bad things. Guess what? They do that with every other kind of statement too. At least half of statisticians' job is fact-checking, and it's a charge we gladly accept.

So the next time you're tempted to say "lies, damned lies, and statistics," or "figures don't lie but liars figure," or "correlation does not imply causation" or any of its variants, or post the umpteen-thousandth link to "How To Lie With Statistics," and think you're being clever--please, just stop. Because one thing I am so sure of that I don't even need to put a p-value on it is that if you feel the need to resort to any of those lazy, thought-free responses, you don't know enough about the issue at hand to have an informed opinion, and the best thing you can possibly do for yourself and everyone else is to keep quiet.

*Opinions vary on this issue, but if statistics isn't exactly a branch of mathematics, we can at least say that math is the language in which it's written.

This discussion has been archived. No new comments can be posted.

Lies, damned lies, and ... oh no, you're going there.

Comments Filter:
  • Liars always lie. I think people mistrust statistics because they don't understand statistics, or worse, understand a little, just enough to be dangerous.

    I worked with data and statisticians my whole career. I'm not a statistician, but learned a lot about the discipline from working with them. One of my co-workers had written a textbook on the subject that was used in colleges. Very interesting discipline.

One small step for man, one giant stumble for mankind.

Working...