(Akaike's information criterion)
Criterion, introduced by Akaike in 1969, for choosing between competing statistical models. For categorical data this amounts to choosing the model that minimizes G2−2ν, where G2 is the likelihood-ratio goodness-of-fit statistic and ν is the number of degrees of freedom associated with the model. An alternative, that usually results in the selection of a simpler model, is the Bayesian information criterion (BIC) for which the quantity minimized is G2−ν ln n, where ln is the natural logarithm and n is the sample size. The latter criterion is also called the Schwarz criterion. A third alternative is the Hannan–Quinn criterion for which the quantity to be minimized is G2−2ν ln(ln n). See also Mallows Cp; stepwise procedures.
Subjects: Probability and Statistics.