correlation vs causation

chrisdavey
chrisdavey Posts: 9,834 Member
edited November 11 in Health and Weight Loss
http://www.soheeleefitness.com/2012/02/29/correlation-vs-causation/

"Don't forget to eat you're protein & lift heavy *kitten*"

GOLD :happy:

Replies

  • taso42
    taso42 Posts: 8,980 Member
    139092366_ce5b410228_o.jpg

    :smile:
  • EricMurano
    EricMurano Posts: 825 Member
    The Correlation vs Causation issue is exactly why we should treat fitness like religion: accept other people's beliefs and don't hold on to your own beliefs when proper research proves you wrong.
  • ATT949
    ATT949 Posts: 1,245 Member
    Thanks for posting this link.
  • joejccva71
    joejccva71 Posts: 2,985 Member
    My name is Joe, and I support the link in the OP's post. Good stuff Chris.
  • chrisdavey
    chrisdavey Posts: 9,834 Member
    Taso: GOLD!

    PU: agreed. But if N=1 & you haven't confirmed that ALL other aspects of your life are the same (which is pretty much impossible for anyone in everyday life IMO) then you can't claim that the run is the single contributing factor to weight loss.
  • Rayman79
    Rayman79 Posts: 2,009 Member
    Phew, I'm glad this was you that posted this Chris. For a second there I thought it was my old Psych research lecturer come back to haunt me! :tongue:

    It's a bit cold here today, I was going to put a jacket on... but now I know I should just hunt and kill a pirate, that'll be much more effective!! :bigsmile:
  • Cytherea
    Cytherea Posts: 515 Member
    As a teacher of philosophy and logic, I just have to say that I :heart: this!
  • Silverkittycat
    Silverkittycat Posts: 1,997 Member
    It's a bit cold here today, I was going to put a jacket on... but now I know I should just hunt and kill a pirate, that'll be much more effective!! :bigsmile:

    What's wrong with pirates? :frown:
  • b2smiley2002
    b2smiley2002 Posts: 19
    Thanks for sharing!
  • Bentley2718
    Bentley2718 Posts: 1,689 Member
    Overall that was a pretty good page, it gets the gist of things right. I would quibble a little about what exactly is necessary to establish causation statistically--you don't absolutely need to be able to control every possible difference, the big thing is just random assignments to conditions so that the counter-factual works, and reasonable controls for other important variables--but this is just me being persnickety. Also, I would say, given the difficulty in using humans in properly controlled experiments for things like diet and exercise programs, I am willing to accept well replicated correlational studies with reasonably good explanations and possibly animal studies. For example, studies linking smoking and cardiovascular problems in humans are entirely correlational, but I still don't think smoking is safe.
  • BrianSharpe
    BrianSharpe Posts: 9,248 Member
    139092366_ce5b410228_o.jpg

    :smile:

    There's a correlation between the number of pirates and global warming; causation is established when we determine that the pirates eat an extremely high fibre diet resulting in frequent emissions of methane - a known greenhouse gas.
  • lipglossjunky73
    lipglossjunky73 Posts: 497 Member
    As a behaviorist who works with teaching people the difference between these 2 things, I give this article a straught up standing O! :drinker:
  • mandasimba
    mandasimba Posts: 782 Member
    Don't forget to eat you are protein and lift heavy ****

    I am protein? Nah, I'm just drunk :p
  • lipglossjunky73
    lipglossjunky73 Posts: 497 Member
    Overall that was a pretty good page, it gets the gist of things right. I would quibble a little about what exactly is necessary to establish causation statistically--you don't absolutely need to be able to control every possible difference, the big thing is just random assignments to conditions so that the counter-factual works, and reasonable controls for other important variables--but this is just me being persnickety. Also, I would say, given the difficulty in using humans in properly controlled experiments for things like diet and exercise programs, I am willing to accept well replicated correlational studies with reasonably good explanations and possibly animal studies. For example, studies linking smoking and cardiovascular problems in humans are entirely correlational, but I still don't think smoking is safe.

    But the variables all need to remain exactly the same except the one you are manipulating specifically...

    I agree that when you use humans, it is near impossible to keep all of the variables fixed....
  • Bentley2718
    Bentley2718 Posts: 1,689 Member
    Overall that was a pretty good page, it gets the gist of things right. I would quibble a little about what exactly is necessary to establish causation statistically--you don't absolutely need to be able to control every possible difference, the big thing is just random assignments to conditions so that the counter-factual works, and reasonable controls for other important variables--but this is just me being persnickety. Also, I would say, given the difficulty in using humans in properly controlled experiments for things like diet and exercise programs, I am willing to accept well replicated correlational studies with reasonably good explanations and possibly animal studies. For example, studies linking smoking and cardiovascular problems in humans are entirely correlational, but I still don't think smoking is safe.

    But the variables all need to remain exactly the same except the one you are manipulating specifically...

    I agree that when you use humans, it is near impossible to keep all of the variables fixed....

    In true experiments, you don't keep things exactly the same (sometimes the things you're studying, e.g. plants, don't even exist before you start), you randomly assign cases to conditions, the idea being that cases have a set (typically equal) probability of ending up in either condition so there should be no systematic differences between the groups (sometimes tests are used to determine whether there are differences between the groups on some set of observables). Based on that, you assume that any differences between the groups post-treatment is due to the treatment. There is a whole field of experimental design, with more complex methods of setting up experiments, but the basic logic remains the same. And a whole field of causal inference. But you probably don't want to visit either.
This discussion has been closed.

Categories