3

I'm selecting models using the information theoretic approach. I've just read that the AICc should be used to rank candidate models where the number of parameters in a model reaches 30% of the sample size. Is this a good benchmark to use when deciding to use AIC or AICc?

luciano
  • 14,269
  • 1
    It would be good to mention/give a link to where you read that. I believe Burnham and Anderson recommend AICc for N-k<40, but I don't have their book handy. – Ben Bolker Feb 16 '14 at 18:15

1 Answers1

5

Burnham and Anderson (see online pdf) recommend using AICc when the ratio n/K is small. In particular, they suggest to use AICc when n/K < 40. In other words, once the number of parameters reach 2% or 3% of the amount of data, you should prefer AICc.

Artem Kaznatcheev
  • 2,649