Frequentists vs Bayesians


But, I mean, c'mon. Frequentists vs Bayesians? Worst. Action. Movie. Ever.

Comments (20)
  1. Anonymous says:

    His follow-up post is funny too:

    The truth is, I genuinely didn’t realize Frequentists and

    Bayesians were actual camps of people—all of whom are now

    emailing me. I thought they were loosely-applied labels—

    perhaps just labels appropriated by the books I had happened

    to read recently—for the standard textbook approach we

    learned in science class versus an approach which more

    carefully incorporates the ideas of prior probabilities.

    >

    I meant this as a jab at the kind of shoddy misapplications

    of statistics I keep running into in things like cancer

    screening (which is an emotionally wrenching subject full

    of poorly-applied probability) and political forecasting.

    I wasn’t intending to characterize the merits of the two

    sides of what turns out to be a much more involved and

    ongoing academic debate than I realized.

    >

    A sincere thank you for the gentle corrections; I’ve taken

    them to heart, and you can be confident I will avoid such

    mischaracterizations in the future!

    >

    At least, 95.45% confident.

    andrewgelman.com/…/16808

  2. Raymond did not insert original title attribute of img. Wow.

  3. Anonymous says:

    Where is the mention of the CC-by-nc license ?

  4. He links directly to the original site. Nitpickers!

  5. Anonymous says:

    @Raphaël

    XKCD's attribution rules:

    "That is, you don't need my permission to post these pictures on your website (and hotlinking with <img> is fine); just include a link back to this page"

  6. Anonymous says:

    Wow, Raymond linking to xkcd. That's a rare one.

    But I too thought this one was really funny and much better than most recent ones. More like the old ones.

  7. Anonymous says:

    I don't get how the neutrino machine is supposed to work.  How can you apply an action (dice roll) to an event (machine telling the truth or not) and say they're related even though they have nothing to do with each other?

    Or do I just not get the joke?

  8. Anonymous says:

    @JJJ: You didn't get the joke.  It runs code that looks like this:

    int neutrino_count = measure_incoming_neutrinos();

    bool sun_has_exploded = (neutrino_count > LOTS_OF_NEUTRINOS);

    if (roll_two_dice() == 36) { sun_has_exploded = !sun_has_exploded; }  // lie

    printf(sun_has_exploded ? "YES" : "NO");

  9. Anonymous says:

    Err, replace that 36 with a 12 (or pretend that roll_two_dice() returns the product of the dice rolls instead of the more traditional sum, but that would just be a wacky API).

  10. sun_has_exploded = !sun_has_exploded

    It's times like this I wish C++ had a zero-ary "!=" assignment operator.

  11. Anonymous says:

    @JJJ. You don't get the joke. The dice-roll is not related to the event. It is related to the truth statement about the event. The joke is that the frequentist notes that the probability of two sixes is less than 5%, and hence concludes that the truth statement yielded "TRUE" with more than 95% certainty, and hence the neutrino device must have predicted "Sun exploded"

    The error, mathematically, is that the error introduced by the two dice dwarfs the likelihood of the event that the truth statement is originally about. So the machine yielding the value "YES" is more likely due to the error introduced by the dice than because the event actually took place.

    The XKCD comic is making a point that given a random variable X with error Y, just because Y < 5% (or any d%) and the random variable yields x, does not mean that the random variable is x with a high or even meaningful confidence. If the error margin is large compared with the variable itself, no meaningful statement can be made until large numbers of readings have been taken place to drive down the standard error of the computation in order to even begin to make meaningful statements about X.

    This is a commentary by the author of XKCD about how bad people are at these kinds of statistics which occasionally crop up in some important situations. If the likelihood that you have breast cancer is small, and the error on the screening is randomly distributed and larger than that value, then just because you've fallen into the category of "machine says you have cancer" doesn't mean you have cancer. It doesn't even mean you're LIKELY to have cancer. It just means that more tests need to be carried out.

  12. Anonymous says:

    @Matt: …just because you've fallen into the category of "machine says you have cancer" doesn't mean you have cancer. It doesn't even mean you're LIKELY to have cancer. It just means that more tests need to be carried out.

    And not only that, but those additional tests sometimes have side effects that may, depending on your personal valuation, be worse than the risk of missing something. Hence the somewhat recent controversial recommendations re. routine breast & prostate cancer screenings, in which doctors no longer encourage it from as early an age in general. (E.g. I think routine mammograms are no longer recommended for 40-50 year olds? I forget. Which isn't to say that it's recommended to NOT get them.) Yes, by skipping them people will be more likely to miss cancer — but they'll also be more likely to skip damaging side effects as well. Where the tradeoff is probably varies significantly from person to person.

  13. Anonymous says:

    @Adam, @Matt:  Ahh, thanks.  I didn't get the joke because I didn't understand the setup.  I get it now.

    Now to tackle that one Far Side comic that I never got…

  14. Anonymous says:

    Worst. Action. Movie. Ever.

    Yeah, well, that's just, like, your opinion, man.

  15. Anonymous says:

    "Frequentists vs Bayesians? Worst. Action. Movie. Ever."

    Got to be better than Ecks vs. Sever

  16. Anonymous says:

    Another interesting tidbit about the comic that someone mentioned in a reply to the post on Andrew Gelman's site linked above: the Sun can't nova; novas only occur on white dwarfs.

    A regular star can supernova but the Sun does not have enough mass for that to occur.

  17. Anonymous says:

    Assume the two stick figures are travelling at constant velocity close to the speed of a neutrino. Would it not be possible that the sun will become a white dwarf and go nova by the time they have finished their conversation?

  18. Mike Dunn says:

    For folks who don't know about explainxkcd: http://www.explainxkcd.com/…/index.php

  19. Anonymous says:

    On the other hand, if you were confident that the Sun had gone nova, you would probably want to spend your last hour as enjoyably as possible, so you might want to take up that bet ;-)

  20. Anonymous says:

    #1 of Internet:

    If any special interest is imaginable, a web site exists for.

    http://www.explainxkcd.made my day!

Comments are closed.