+ Reply to Thread
Results 1 to 3 of 3

Thread: AIC backward selection - control variables

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

    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    AIC backward selection - control variables




    Hej!

    I have a question concerning AIC backwards selection. I am doing a Linear Mixed effect model where I have one main predictor and several control variables (covariates).
    I did AIC backward selection and removed the not-significant control variables. Now I have the following situation: my main predictor is NOT significant. But two of the control variables are significant. So how do I continue the model selection? Can I remove the control variables even though they are significant? If I do so, and I only leave the main predictor in the model, the AIC does get smaller but the main predictor is still not significant. But is that allowed?
    I am very confused how to deal with the model now. And I am very thankfull for any help!

    Best regards,
    P

  2. #2
    Devorador de queso
    Points: 95,819, Level: 100
    Level completed: 0%, Points required for next Level: 0
    Awards:
    Posting AwardCommunity AwardDiscussion EnderFrequent Poster
    Dason's Avatar
    Location
    Tampa, FL
    Posts
    12,935
    Thanks
    307
    Thanked 2,629 Times in 2,245 Posts

    Re: AIC backward selection - control variables

    If you have control variables don't you want them in the model so that you can control for them? It sounds like your confusion just arises since the outcome isn't quite what you were hoping for.
    I don't have emotions and sometimes that makes me very sad.

  3. #3
    Omega Contributor
    Points: 38,374, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    6,998
    Thanks
    398
    Thanked 1,186 Times in 1,147 Posts

    Re: AIC backward selection - control variables


    If the primary objective was to examine the affect of the main term, well keep it in the model along with the control variables and report the null effect.


    Also, if you are miffed by the outcomes, you may also want to look at the effect size of the main term to see if it truly is small or if your project may just be underpowered.

+ Reply to 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