+ Reply to Thread
Results 1 to 2 of 2

Thread: How to use Akaike's Information Criterion (AIC) for manual model reduction?

  1. #1
    Points: 497, Level: 9
    Level completed: 94%, Points required for next Level: 3

    Posts
    6
    Thanks
    0
    Thanked 0 Times in 0 Posts

    How to use Akaike's Information Criterion (AIC) for manual model reduction?




    Hi all,

    I wondered if anybody can explain to me how I can use the AIC for manual reduction of my GLM. I calculate a GLM with the following independent variables (full model):

    a
    b
    c
    d
    a*b
    a*c
    a*d
    b*c
    b*d
    c*d
    a*b*c
    a*b*d
    a*c*d
    b*c*d

    Some of those main effects/interactions went significant, others don't. Is there any stepwise reduction procedure (based on the AIC) I can follow to skip some of the not-significant main effects/interactions?

    I already did a quick web-search and always found the information that I can use the AIC to compare two (or more) models (the one with the lowest AIC is the better one) but I am looking for a procedure to reduce my full model similar to the classical stepwise backward reduction procedure based on p-values.

    Please don't just suggest R commands like step() or stepAIC(). That would not help since I'm not working with R. But it would probably help if you know what those commands do. Maybe I can perform those procedures manually in my statistic-software.

    Thanks!

    Fred.

  2. #2
    Points: 4,664, Level: 43
    Level completed: 57%, Points required for next Level: 86
    kiton's Avatar
    Location
    Corn field
    Posts
    234
    Thanks
    47
    Thanked 51 Times in 46 Posts

    Re: How to use Akaike's Information Criterion (AIC) for manual model reduction?


    Hello Fred,

    Is your goal to get rid of some interactions? If so, there are two comments here:

    (A) If your theory says an interaction should be present -- leave it there even if it is non-significant;
    (B) If you still want to reduce the number of interactions, consider the guidelines by Aiken, L. S., West, S. G., & Reno, R. R. (1991). Multiple regression: Testing and interpreting interactions. Sage. and exclude those interactions (1) that are non-significant and (2) for which the percent of explained variation (delta R-squared) is non-significant.

    As for using AIC, the model with lower value indicates a better fit. As such, if removal of an interaction increases the AIC, then you better keep it and vice versa.

+ 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