Which statement about using AICc is correct?

Which statement about using AICc is correct?

• A. Use AICc when the sample size is greater than 200.
• B. Use AICc when the sample size is under 200.
• C. Use AICc when the sample size is under 50.
• D. Use AICc for all sample sizes.

Answer: (B)

AICc is the small-sample corrected version of AIC, designed to guard against choosing overly complex models when data are limited. The correction adds a penalty that grows with the number of estimated parameters and shrinks as the sample size increases. When the sample size is small relative to the number of parameters, this correction is sizeable, pushing the selection toward simpler models. As the sample size gets large, the correction term becomes negligible and AICc behaves like AIC, so the distinction matters mainly in small samples. So, the statement describing using AICc for smaller samples aligns with this idea: you apply the correction when the data are not ample compared to model complexity. The other options imply using AICc in situations where it isn’t necessary (or always), which isn’t correct because the need for the correction diminishes with larger data.

AICc is the small-sample corrected version of AIC, designed to guard against choosing overly complex models when data are limited. The correction adds a penalty that grows with the number of estimated parameters and shrinks as the sample size increases. When the sample size is small relative to the number of parameters, this correction is sizeable, pushing the selection toward simpler models. As the sample size gets large, the correction term becomes negligible and AICc behaves like AIC, so the distinction matters mainly in small samples.

So, the statement describing using AICc for smaller samples aligns with this idea: you apply the correction when the data are not ample compared to model complexity. The other options imply using AICc in situations where it isn’t necessary (or always), which isn’t correct because the need for the correction diminishes with larger data.