Akaike Information Criterion (AIC)
and the Schwarz Information Criterion
*************************************************** * The Akaike Information Criterion compares an * * improvement (increase) in max log(likelihood) * * against added complexity of the model caused * * by additional parameters (i.e. overfitting). * *************************************************** max log(likelihood): without POD floor or ceiling = -230.49 with POD ceiling only = -206.541 *** Larger maximum likelihood is better. Akaike Information Criterion (AIC): without POD floor or ceiling = 464.979 with POD ceiling only = 419.083 *** Smaller AIC is better. ************************************************ Based on the AIC, using a POD ceiling model appears to be justified. ************************************************ The Schwarz Criterion (Bayesian Information Criterion) is more conservative than the AIC. Schwarz Criterion (BIC): without POD floor or ceiling = 473.045 with POD ceiling only = 431.182 *** Smaller BIC is better. Warning: Floor/Ceiling models should be based on a minimum of 150 observations. Models with Floor AND Ceiling should be based on a minimum of 300 observations. Number of observations in this analysis: 417 MIL-HDBK-1823, v7.1.6 (WorkshopMenu, v3.1) ********************************************