Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • AIC and BIC model selection

    Do we use BIC, instead of AIC, when AIC is unstable?

    I ran -varsoc x, maxlags(5)-, and both AIC and BIC showed 5 lags.
    However, when I increase the maxlags(), the BIC won't change, and still show 5 lags, while AIC will change for each maxlags() selection. There seems to be a pattern, and AIC will pick n-3 lags, where n is the maxlags(). So if I do -varsoc x, maxlags(30), AIC will choose 27 lags.

    Does this make sense? Should I trust BIC and use 5 lags, as it will consistently pick 5 lags regardless of maxlags I use.

    Thanks

  • #2
    hi sir, Give me the mathematical formula for this commends in stata rnormal and runiform

    Comment


    • #3
      How to count aic and bic with stata

      Comment


      • #4
        Originally posted by retia mohamed View Post
        hi sir, Give me the mathematical formula for this commends in stata rnormal and runiform
        Hi Retia. I don't have any particular mathematical formula, I just use Stata for this task.

        Comment


        • #5
          If you look for 5 minutes in the documentation (particularly the subject index) you'll find the entries on BIC and AIC. These entries will tell you how Stata calculates AIC and BIC and what to type in to get them. varsoc does these for you. Also, any good econometrics text will cover these. Generally, you're much better off using Stata routines than programming things yourself.

          There are debates in the literature over AIC versus BIC. One has a larger correction for parameters than the other and so favors smaller models.

          Comment


          • #6
            thank you

            Comment


            • #7
              Originally posted by Phil Bromiley View Post
              If you look for 5 minutes in the documentation (particularly the subject index) you'll find the entries on BIC and AIC. These entries will tell you how Stata calculates AIC and BIC and what to type in to get them. varsoc does these for you. Also, any good econometrics text will cover these. Generally, you're much better off using Stata routines than programming things yourself.

              There are debates in the literature over AIC versus BIC. One has a larger correction for parameters than the other and so favors smaller models.
              Hi. I did it, but I don't understand why lag order for AIC changes for each maxlag.

              Comment

              Working...
              X