1. Exquisite Tweets from @rabaath, @rlmcelreath, @omaclaren, @l__ds

    Woody_WongECollected by Woody_WongE

    My 30 word #BetterBayes entry:

    Use the language of probability to describe what you know and what is uncertain about a situation. Add data. Voilà! Bayes uses this new information to optimally reduce that uncertainty.

    Contest: statsandstories.net/betterbayes/ from @statsandstories

    Reply Retweet Like

    rabaath

    Rasmus Bååth

  2. Bayesian Analysis Means Counting Possibilities

    How compatible are different ideas with data? Count the ways the data could occur, according to each idea. Ideas with more ways to produce the data are more plausible. @statsandstories #BetterBayes

    Reply Retweet Like

    rlmcelreath

    Richard McElreath

    I blame @rabaath for making me do this. twitter.com/rabaath/status…

    Rasmus Bååth @rabaath
    My 30 word #BetterBayes entry:

    Use the language of probability to describe what you know and what is uncertain about a situation. Add data. Voilà! Bayes uses this new information to optimally reduce that uncertainty.

    Contest: statsandstories.net/betterbayes/ from @statsandstories

    Reply Retweet Like

    rlmcelreath

    Richard McElreath

  3. I love this description, but then I would also love to have a straight answer to those pesky "continuoist" who will inevitable ruin the party by asking "So what about continuous probability? Have you heard about measure theory?" . Is there a way to gracefully dismiss them?

    Reply Retweet Like

    rabaath

    Rasmus Bååth

  4. "Die ganzen Zahlen hat der liebe Gott gemacht, alles andere ist Menschenwerk!"

    But seriously, I don't dismiss. Counting in continuous spaces is tricky, no way out of that. No unique construction, cf non-standard analysis.

    Anyway, metaphors have limits.

    Reply Retweet Like

    rlmcelreath

    Richard McElreath

  5. I think I'll stick to saying that quote and when they have finished looking that up on Google translate I'm long gone! :)

    Not dismissing, but just standing a chance against the argument that measure theoretical probability is the only correct definition.

    Reply Retweet Like

    rabaath

    Rasmus Bååth

  6. Missing tweet: 1041973092635926528

  7. My admittedly annoying response is that this describes likelihood really well but not really probability...Probability adds eg that only one idea is correct (a ‘fixed probability pie’ to allocate among ideas)

    Reply Retweet Like

    omaclaren

    Oliver Maclaren

  8. s/Bayesian Analysis/likelihood/

    —or am I missing something? For Bayes, don't you also need to count the ways the ideas could be true?

    Reply Retweet Like

    l__ds

    l__ds

  9. Short answer: No, truth is for philosophers.
    Less short answer: Ideas need initial weights, just like in every other paradigm. Bayes allows those initial weights to differ. But it is still just counting.

    Reply Retweet Like

    rlmcelreath

    Richard McElreath

    Probability requires normalization. But Bayes works without normalizing. Sum and product rules don't get their power from normalization, but rather from counting possibilities.

    Reply Retweet Like

    rlmcelreath

    Richard McElreath