+ Reply to Thread
Page 2 of 2 FirstFirst 1 2
Results 16 to 24 of 24

Thread: Postive correlation but negative coefficient in regression??

  1. #16
    TS Contributor
    Points: 22,359, Level: 93
    Level completed: 1%, Points required for next Level: 991
    spunky's Avatar
    Location
    vancouver, canada
    Posts
    2,135
    Thanks
    166
    Thanked 537 Times in 431 Posts

    Re: Postive correlation but negative coefficient in regression??




    Quote Originally Posted by jpkelley View Post
    Interesting, spunky. Good call about the meta-analysis. What fields did you include...human behavioral sciences only, or did you delve into the ecology and animal behavior fields as well?
    only human behaviour stuff... the other 1/2 of my academic training is in psychology (and this friend's in philosophy) so we have to keep it within areas we understand so we can make sense of whatever people are interpreting... but that quote from Paulhus' article is what kind of prompted this whole research thing and it's like opening a pandora's box... suppression is really out there and it really happens a lot more often that people give it credit for, but i dont really know why most social sciences researchers either choose to ignore it completely (most common solution), blame it on sampling error or present it on the "future research is needed" section... very, very few people actually id it as suppression and, those who do so, do it kind of like "ok, so we have a suppressor variable there. moving on to the F test, we can see that blah, blah, blah..."
    for all your psychometric needs! https://psychometroscar.wordpress.com/about/

  2. #17
    Cookie Scientist
    Points: 13,431, Level: 75
    Level completed: 46%, Points required for next Level: 219
    Jake's Avatar
    Location
    Austin, TX
    Posts
    1,293
    Thanks
    66
    Thanked 584 Times in 438 Posts

    Re: Postive correlation but negative coefficient in regression??

    Sounds like a very interesting paper, spunky. As you can see, I'm as guilty as the next behavioral researcher in regarding suppression effects as quirky oddities...

  3. #18
    Points: 7,911, Level: 59
    Level completed: 81%, Points required for next Level: 39

    Posts
    39
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Re: Postive correlation but negative coefficient in regression??

    Hi all,
    I'm glad that the results I've presented have stimulated some interesting conversation and have highlighted some important considerations. I now understand the results I am getting and due to the nature of my variables can see how one is 'suppressing' the other.
    However I am still none the wiser on how to report this.

    I have reported the positive correlation between predictor 1 and outcome, and the negative correlations between predictor 2 and ouctome. Can I then say in my regression write up something along the lines of:

    'Despite the negative correlation observed between predictor 2 and the outcome, there is evidence of a strong suppression effect with the results of the multiple regression indicate that both predictors are postively related to the outcome variable.'

    and leave it at that? (with further explanation in the discussion)

  4. #19
    TS Contributor
    Points: 22,359, Level: 93
    Level completed: 1%, Points required for next Level: 991
    spunky's Avatar
    Location
    vancouver, canada
    Posts
    2,135
    Thanks
    166
    Thanked 537 Times in 431 Posts

    Re: Postive correlation but negative coefficient in regression??

    Quote Originally Posted by statto View Post
    (with further explanation in the discussion)
    as long as you elaborate on the suppression (which is kind of the most interesting part of finding it), i think it sounds pretty ok...
    for all your psychometric needs! https://psychometroscar.wordpress.com/about/

  5. #20
    Points: 199, Level: 3
    Level completed: 98%, Points required for next Level: 1

    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Postive correlation but negative coefficient in regression??

    Do your predictor variables and your outcome variables have to be correlated in the original bivariate analsis in order to do a single or multiple regression between the two?

  6. #21
    Points: 32, Level: 1
    Level completed: 64%, Points required for next Level: 18

    Location
    Lagos, Nigeria
    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Postive correlation but negative coefficient in regression??

    Hello all.

    I just joined this forum, and find it very interesting.

    I have a similar experience with my regression analysis, except that there are three predictor variables in mine. I have not had much experience with suppression effect before now. Your explanations to Statto have been of great help.

    Thank you.

  7. #22
    Fortran must die
    Points: 58,790, Level: 100
    Level completed: 0%, Points required for next Level: 0
    noetsi's Avatar
    Posts
    6,532
    Thanks
    692
    Thanked 915 Times in 874 Posts

    Re: Postive correlation but negative coefficient in regression??

    My sense is that you can not tell anything about the multiple regression coefficients (that use partial slopes) from the bivariate relationships. Sometimes they will be similar sometimes entirely different
    "Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995

  8. #23
    Points: 4, Level: 1
    Level completed: 7%, Points required for next Level: 46

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Postive correlation but negative coefficient in regression??

    Thankx Jake for this explanation
    Really it's Great

  9. #24
    Points: 2, Level: 1
    Level completed: 3%, Points required for next Level: 48

    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Cool Re: Postive correlation but negative coefficient in regression??


    hi.

    i am facing the same situation right now. the univariate analysis for the three variables were positive, however when conducting multiple regression in multivariate anaysis, these variables became negatively associating with the outcome. my questions are:
    1. how to identify which variables suppressed
    2. how to rectify this problem from the spss output
    3. what should we do about it

    i am dealing with overload of data (national health registry) witb a total of 6k patients.

    thanks

+ Reply to Thread
Page 2 of 2 FirstFirst 1 2

           




Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats