Guess the correlation

Discussion in 'Research methodology news and research' started by ME/CFS Skeptic, Jul 20, 2021.

  1. ME/CFS Skeptic

    ME/CFS Skeptic Senior Member (Voting Rights)

    Messages:
    4,002
    Location:
    Belgium
    Here's a fun, stupid game: guess the correlation.
    http://guessthecorrelation.com/

    A random scatterplot is given and you have to guess the correlation coefficient (a number between 0 and 1 that indicates the strength of relationship between the two variables).

    My high score is 155 points Who can do better?
     
  2. ME/CFS Skeptic

    ME/CFS Skeptic Senior Member (Voting Rights)

    Messages:
    4,002
    Location:
    Belgium
  3. Trish

    Trish Moderator Staff Member

    Messages:
    55,414
    Location:
    UK
    That was fun. My score improved from 33 to 104 on my first 2 goes.
     
    Simbindi, DokaGirl, Louie41 and 3 others like this.
  4. Dolphin

    Dolphin Senior Member (Voting Rights)

    Messages:
    5,795
    high score.PNG
    My scores though didn't increase monotonically

    108
    1 (oops!)
    175
     
  5. Woolie

    Woolie Senior Member

    Messages:
    2,922
    Haha, nice way of learning what the patterns look like in terms of numbers. Its a really good demonstration of how one or two outliers can really skew the value of r.

    I couldn't get the game to stop, so I just quit out after a while.
     
  6. Trish

    Trish Moderator Staff Member

    Messages:
    55,414
    Location:
    UK
    It stops when you lose all 3 hearts by getting 3 badly wrong in quick succession.
    As well as showing the effect of outliers, I think it also illustrates how similar the patterns look in the bottom half of the range, so a correlation of 0.4 looks to me very much like a correlation of 0.1. And to me 0.4 often looks much like 0.6. This probably partly indicates that my eye naturally eliminates outliers and focuses on the mass in the middle, and that the formula gives too much weight to outliers, or it could suggest that the eye deceives and I need the formula to help spot subtle trends.

    My inclination is to the former. If I can't see a definite trend visually in the whole mass of data, not just a few outliers, then in the real world there probably isn't one of clinical significance, and if a correlation is marginally statistically significant and a correlation is being claimed, then perhaps the data is being over interpreted to suit the researcher's purposes.
     
    Simbindi likes this.
  7. ME/CFS Skeptic

    ME/CFS Skeptic Senior Member (Voting Rights)

    Messages:
    4,002
    Location:
    Belgium
    Evelien and I wanted to get 1000 points but we didn't make it.
    upload_2021-7-21_14-55-56.png
     
  8. Snow Leopard

    Snow Leopard Senior Member (Voting Rights)

    Messages:
    3,860
    Location:
    Australia
    Looking at the high scores (on the site), I think some people are testing algorithms... No way any real person would have that sort of patience.
     
    Simbindi, Trish and ME/CFS Skeptic like this.
  9. Keela Too

    Keela Too Senior Member (Voting Rights)

    Oh my! I thought I would be better at this game! :unsure:

    [Edit to add: Okay, bit of practice, and I’ve now managed 204. Ho hum!]
     
    Last edited: Jul 22, 2021
    Simbindi, Trish and ME/CFS Skeptic like this.
  10. TiredSam

    TiredSam Committee Member

    Messages:
    10,557
    Location:
    Germany
    upload_2021-7-30_0-1-5.png

    Rubbish game.
     
  11. Jaybee00

    Jaybee00 Senior Member (Voting Rights)

    Messages:
    2,177

Share This Page