Friday, March 6, 2009

Comic 552: Causing Problems

more like correlaSHIT

Well, looks like someone has been taking statistics classes. What we have is a basic little statistics joke. He is noticing a correlation about correlations! and he is wondering if learning about causation has caused him to...um...think that way about causation I guess. It's a simple joke, told in simple pictures, and I suppose it's clever for what it is. I can't get very excited about it (it's not as clever as the last statistics comic) but I can't say it's terrible either. Certainly better than memes + wikipedia, and nerdy in the way xkcd used to be. This one totally feels like he could have written it in his notebook in stat class.

A different angle: like with some previous comics, this one seems like a general attempt to make an easy joke that stat people will like, print, cut out, and put on their walls. "A special xkcd, just for me!" they will think. Please: do not blindly e-mail this comic to all of your friends who use statistics. It is annoying. It is not nearly funny enough, or faliing that obscure enough, to warrant spreading like beer pong herpes.

---------

My heartiest laugh that was correlated to (and caused by) a comment about correlation and causation was when Nate Silver was debating John Zeigler on some dumb online tv-type show. The conversation went something like this:

JZ: But 75% of the people who heard [X] voted for [Y]!
NS: Ok, but that doesn't mean they voted for [Y] because they heard [X]. You of all people should know that just because two data points are correlated doesn't mean one caused the other.
JZ: But with numbers as high as 75% ??? Come on!

This made me laugh, because JZ was being a giant idiot. You can have as much correlation as you want and it still can mean nothing. ha ha, stupid people.

----------

From nowhere: I find that God's eyes blinking on this Overcompensating freaked me out. A comic like that, that is just about always static, should stay static. This is crotchety-old-man me speaking, but I just think it's weird. (this dinosaur comic is a little better because the movement is so obvious, it doesn't sneak up on you all of a suddent)


======
update: See how up there I predicted that statisticsy people would spread this comic around to all their friends because hey, look, it's a statistics comic! ? Yeah well I was at least a little right: Noted smart person Steven Levitt has apparently broken his "no cartoons" rule for the very first time to bring you you-know-what. curse you, levitt!

41 comments:

  1. What did you mean by "faliing that obscure enough"?

    Anyway agreed with the general sentiment. But it's definitely one of the better xkcds lately. Not wonderful, but okay.

    ReplyDelete
  2. Wait nevermind. Pretend I changed my comment to say that you have a typo.

    ReplyDelete
  3. Thank you for your blog.

    This comic is really lame and I'm a bit sick of seeing it everywhere. Every once in a while, it marginally doesn't suck.

    ReplyDelete
  4. Agreed, this one was pretty damn decent.

    Now bring on the cuddlefish whining about how xkcdsucks is posting POSITIVE STUFFS OH NO.

    ReplyDelete
  5. Is the purpose of this blog to point out when XKCD sucks, or to criticize every single comic released by it? Because I think you might make a more forceful impact if you didn't have an entry about every single XKCD that comes out, whether it sucks or not.

    I agree that XKCD is hit or miss, and while it may be very popular, it has generally lost it's zing over time.

    As an example, it seems like your primary criticism for 552 is that it seems to you that Munroe is purposefully trying to make it onto refriderators and bulliten boards via... a "simple, clever for what it is, joke".

    I like the idea of an XKCD Sucks blog existing, but I think the lack of focus on the truly sucky ones weakens it.

    ReplyDelete
  6. Yay, a comic that could have been one of the average ones from a few years ago. A little forced, but a cute, geeky chuckle as is befitting.

    ReplyDelete
  7. I don't hate this one.

    My world is falling apart.

    ReplyDelete
  8. I took a statistics course last semester, but I didn't need that to find this joke trite and uninspired.

    ReplyDelete
  9. I agree, good comic.

    But that middle section of your analysis is HURTING for lack of context.

    ReplyDelete
  10. I suppose I could be making the correlation/causation mistake myself, but i think you have a soft spot for statistics-related xkcds. This one does suck, and so did the last one. Utterly predictable.

    ReplyDelete
  11. I thought it was so-so. Not bad, but not good either.

    ReplyDelete
  12. spockodile, you said, "I think the lack of focus on the truly sucky ones weakens it."

    I understand your point about how Carl's entries would be sharper if he only hated on the worst entries, but I think there's currently a healthy focus on the worst offenders. When Randall truly screws the pooch, the comments page will begin preliminary groaning and the update about the entry will often be ranty and angry.

    ReplyDelete
  13. I agree, this one is quite funny. Hurray for Randall- now home for tea and scones.

    ReplyDelete
  14. "You can have as much correlation as you want and it still can mean nothing."

    Correlation does NOT mean 'nothing'. It just does not mean causation.

    ReplyDelete
  15. Eh.

    In other news: HOW DID I MISS THAT INTERVIEW DEBATE THING THAT IS THE BEST THING EVER

    ReplyDelete
  16. this is officially the stupidest blog I've ever read

    ReplyDelete
  17. That says more about you than it does about the blog, Cuddlefish my dear.

    ReplyDelete
  18. Any work on that "forms of cuddlefish" page yet?

    I think most of the reason I liked this one is because I don't have to think about the previous one anymore. I'll never be able to use an Etch-a-Sketch again.

    ReplyDelete
  19. "Correlation does NOT mean 'nothing'. It just does not mean causation."

    "You can have as much correlation as you want and it still CAN mean nothing."

    Grammar.

    ReplyDelete
  20. I should probably solicit posts on that, it will make a good communal project.

    ReplyDelete
  21. Feel free to take charge; it's a little to introspective for me. But go for it. Guys do you see the new "Meta" links on the left panel there? These make me very happy.

    ReplyDelete
  22. lol Carl.

    Oh oh Thomas Thomas I forgot to tell you that when you guest posted I told a friend of mine about this site and said that your complaints were legit. He still likes xkcd but admits that it is not so good lately. ONE PERSON DOWN, a bajillion more to go!

    ReplyDelete
  23. Amanda: Thank you for sharing that. I've only ever mentioned xkcdsucks in a handful of actual conversations, but the moment I do I discover the other person corresponds snarky blogs with drowning puppies. "Why the HELL would anyone do that?!"

    ReplyDelete
  24. ""Correlation does NOT mean 'nothing'. It just does not mean causation."

    "You can have as much correlation as you want and it still CAN mean nothing."

    Grammar."


    Uh, yes? I fucking read that. Do you think so little of us anonymous posters that you jump to ridiculous conclusions? Asshole.

    How about YOU read what I said? "Correlation does not mean nothing" ie. "it CANNOT mean nothing".

    Any amount of information means something, regardless of what it is, even if it's a lie. Hell even NO information is meaningful.

    ReplyDelete
  25. "Any amount of information means something, regardless of what it is, even if it's a lie. Hell even NO information is meaningful."

    Only operating from the axiomatic assumption that information (as we perceive it) actually exists, and that meaning is not an imagined, metaphysical by-product of our ability to reason.

    Language is awful at encoding meaning, and even more awful at conveying it.

    Also, I think the context of the original statement implied that correlation can provide data that, while it inherently has what we perceive to be meaning, does not have any useful meaning.

    If you want textual language to always be context free, then your complaint is understandable. Otherwise, learn to interpret contextual clues and stop being such an ass about semantics.

    ReplyDelete
  26. Asshole Anon: Correlation does not mean nothing. Days that you leave comments on my blog are correlated to days when I eat pizza. This means nothing, as the two are not connected, are not both caused by some third factor, or anything like that.

    I think little of you not because you are anonymous - lots of anonymous posters manage to write intelligent, reasoned things - but when you write something dumb, are called on it, and then defend yourself with "well i do know grammar SO BOO YAH" instead of defending your point - that correlation means something (or as you put it, "does not mean nothing") when that's clearly not always true.

    ReplyDelete
  27. I did laugh at this one, in a good way, so huzzah!
    But he'll probably go and screw it all up with the next one...just like he did last time with that huge plethora of wikipedia tripe.

    ReplyDelete
  28. Surely he has causation - instead of saying maybe he should say "yeah". He took a stats class and it lead to more knowledge - that's causation - maybe is wrong!

    ReplyDelete
  29. Also I feel like the alt-text is very DC-esque.

    ReplyDelete
  30. "Days that you leave comments on my blog are correlated to days when I eat pizza. This means nothing, as the two are not connected, are not both caused by some third factor, or anything like that."

    Again... it doesn't 'mean nothing' it just doesn't mean that they have a causal connection. It definitely does NOT mean that they don't have a causal connection.

    Not meaning one thing does not mean meaning NOTHING.

    "as the two are not connected, are not both caused by some third factor, or anything like that."

    How could you possibly know that?

    ""well i do know grammar SO BOO YAH" instead of defending your point""

    Because he didn't make a point against me at all? All he said was that I didn't understand grammar and I replied showing I did entirely understand what was said.

    Where is your problem with that? Your problem with that is artificially constructed to defend your ridiculous claim and fight the 'asshole cuddlefissh' that you need to show a front against.

    "Only operating from the axiomatic assumption that information (as we perceive it) actually exists"

    The ontology of information doesn't matter, we can only deal with things as we perceive.

    "and that meaning is not an imagined, metaphysical by-product of our ability to reason."

    Again, the ontology doesn't matter. Whether meaning is an intrinsic property or not is irrelevant. We can only deal with what we perceive and understand. So such ontological axioms are ridiculous to even bother justifying. They're useless in the end.

    "I think the context of the original statement implied"

    Implication and intent are worse evidential points than correlation. Much worse.

    "does not have any useful meaning."

    I'm glad you made this distinction since it's in essence what whatshisface said wrong. Now I don't need to go and tell him why he is retarded because you did it for me. 'Meaningful' =/= 'useful' and never will be.

    ReplyDelete
  31. I see some serious beard stroking going down.

    ReplyDelete
  32. "Whether meaning is an intrinsic property or not is irrelevant. We can only deal with what we perceive and understand. So such ontological axioms are ridiculous to even bother justifying. They're useless in the end."

    Useless, but not necessarily meaningless ;)

    Hurrah for strange loops!

    "Implication and intent are worse evidential points than correlation. Much worse."

    If this were a legal case, yes. If it's just a conversation, the inability to infer implication from contextual clues indicates a lack of conversational experience, deficient social skills, or a huge cultural divide between the speakers.

    ReplyDelete
  33. Anonymous is right. Correlation often does mean something. All information that we have on the connection of smoking to lung cancer in humans is of a correlative nature (ethics committees tend to look favorably on people being assigned to ingest possibly toxins in a 30+ year longitudinal study).

    In fact, the Surgeon General put the warning label that we currently have on cigarettes due to there being so many correlative studies with various variables that it was deemed sufficient for a causal link.

    There are many fields of research where correlative data is the best we can possibly get due to technical limitations and ethics. However, we should still study these areas, realizing that there is a caveat to correlative data and keeping our minds open to other theories that explain the relationship between two variables.

    To see my point, it may help to use an actual example that is commonly used in stats/methodology courses (rather than a silly made up one). There is an extremely high correlation between shark attacks and sales of ice cream. There are three options here. We can either write this off as you would do, scoffing. We could assume that ice cream somehow attracts the sharks (or that people buy ice cream to reduce swell after being attacked by a shark). The latter option seems absurd and would lead us aboard your train of thought.

    There is a third option though. Hot temperatures increase both the sale of ice cream AND the amount of people swimming in the shark. Although we still haven't "proven" anything, from what we know about animal behavior this seems quite likely. There was an important connection between the two that we would have missed entirely.

    Correlative data can be good for research in that it can give a pretty good idea as to what to study in the future. As I said, sometimes there are ethical/technical issues that are prohibiting research. However, new advances (in technology, not ethics) can allow us to do experimental research on these things in the future, and we'll have a fairly decent idea as to what to look for at that time thanks in part to correlative data.

    That said, I get annoyed as well when I see correlative misrepresented in the media. Correlative data is very important, but we shouldn't stop there.

    ReplyDelete
  34. I'd always heard at as bikinis and algae. Evidently my stats class focussed more on plants and girls than food and fish.

    But yeah, correlation in general means that it is very likely one causes the other or some third factor causes both. It is of course possible that they are unrelated but by chance, but it is possible that if I drop an egg from fifty feet it won't break (it could land on a slowly decelerating bird, for instance). It can mean nothing, true, but anything can mean nothing. It probably doesn't mean nothing.

    ReplyDelete
  35. Wow, that looked a lot more impressive in comment box. Damn font sizes.

    ReplyDelete
  36. Poore, how can we possibly know the internal state of the writer?

    Fake Smile, if you're a projectivist then correlation is the best you can get.

    ReplyDelete
  37. Translation of this comic:

    HAI GUYZ! I not only know common 4chan memes, I ALSO know common slashdot memes!!

    ReplyDelete
  38. Uh... that's not really a slashdot meme unless slashdot has really changed a lot since I was there last (about a year ago).

    I mean yeah, people said it there sometimes, but that doesn't necessarily make it a "slashdot meme." I would say that it's a meme that runs through western culture, especially in the scientific community.

    ReplyDelete
  39. Steven D. Levitt, one of the co-authors of the book Freakonomics, liked this comic and essentially printed, cut out, and put this on his internet wall, the Freakonomics blog. Someone apparently e-mailed it to him.

    Good call.

    ReplyDelete
  40. man i know i am like the saddest person in ever for reading old xkcdsucks posts, but still

    from the freakonomics blog:

    "Thanks for reposting this XKCD joke for all of your Freakonomics readers.

    This is one of the funniest cartoons that I’ve seen in a long time and the esoteric punchline adds to the allure. I’ve shown it to some non-stats geeks and all I get is blank stares."

    tl;dr "I get this and some other people don't! This means it's HILARIOUS!"

    Reading that made me sadface :(

    ReplyDelete
  41. i know. it's why people who say "just don't read it if you don't like it" are wrong. Alas, the desire to see esoteric things that you understand is strong, even for the freakonomics folks.

    nothing wrong with posting on old posts, i am much more likely to reply that way...

    ReplyDelete