Interpreting AIC in @RISK

Applies to: @RISK 6.x+, Professional & Industrial Editions

Overview:
@RISK shows candidate distributions with AICc values (displayed as AIC). Differences in AICc (Δ) can be used to compute evidence ratios, showing relative likelihoods of fits.

Example:

• Normal fit AICc = –110, triangular fit AICc = –106 → Δ = 4

• Evidence ratio ≈ 7.4 → normal is 7.4× more likely than triangular

Guidelines:

• Δ = 2 → ratio ≈ 2.7

• Δ = 8 → ratio ≈ 54.6

• Higher Δ → stronger evidence against a fit, but no fit can be rejected with certainty.

• Always check fitted curves vs. your data histogram; a lower-ranking fit may be better in regions that matter.

Reference:

• Anderson, Model Based Inference in the Life Sciences, Springer, 2008

• Burnham et al., Behav Ecol Sociobiol (2011) 65:23–35