
Model Selection Criteria 127
where the last term on the right-hand side of the above equation is often
referred to as the bias-correction term. We observe that as the sample size
N → ∞ this bias-correc tion term approaches zero, and the resultant criterion
is just the usual AIC.
It should be noted that the bias-correctio n term in (4.13) changes if a dif-
ferent probability distribution (e.g., exponential, Poisson) is assumed for the
measurement errors. However, it was suggested in [17] that in practice AIC
c
given by (4.13) is genera lly suitable unless the underlying probability distri-
bution is extremely non-normal, espe cially in terms of being