Compare AIC Using PROC MIXED -------------------------------- AIC: -2*log(Likelihood(hat(beta))) + 2p BIC: -2*log(Likelihood(hat(beta))) + p*log(N) Example 1 ------------ DATA SET1; INPUT GENDER $ HEIGHT WEIGHT AGE; /* Compare AIC*/ HEIGHT2=HEIGHT**2; DATALINES; M 68 155 23 F 61 99 20 F 63 115 21 M 70 205 45 M 69 170 . F 65 125 30 M 72 220 48 ; ###Only Height PROC MIXED DATA=SET1; MODEL WEIGHT=HEIGHT; RUN; -2 Res Log Likelihood 45.4 AIC (smaller is better) 47.4 AICC (smaller is better) 48.8 BIC (smaller is better) 47.0 ###HEIGHT and HEIGHT**2 PROC MIXED DATA=SET1; MODEL WEIGHT=HEIGHT HEIGHT2; RUN; -2 Res Log Likelihood 42.2 AIC (smaller is better) 44.2 <----Smaller AICC (smaller is better) 46.2 BIC (smaller is better) 43.6 <----Smaller --------------------- Example 2 ------------ ### HR=a+b*DOSE+e DATA HEART; INPUT DOSE HR; DATALINES; 2 60 2 58 4 63 4 62 8 67 8 65 16 70 16 70 32 74 32 73 ; PROC MIXED DATA=HEART; MODEL HR=DOSE; RUN; -2 Res Log Likelihood 44.6 AIC (smaller is better) 46.6 AICC (smaller is better) 47.2 BIC (smaller is better) 46.7 ### HR=a+b*log(DOSE)+e, <--Dose on log-scale. DATA HEART; INPUT DOSE HR; LDOSE=LOG(DOSE); DATALINES; 2 60 2 58 4 63-2 Res Log Likelihood 23.7 AIC (smaller is better) 25.7 AICC (smaller is better) 26.4 BIC (smaller is better) 25.8 4 62 8 67 8 65 16 70 16 70 32 74 32 73 ; PROC MIXED DATA=HEART; MODEL HR=LDOSE; RUN; -2 Res Log Likelihood 23.7 AIC (smaller is better) 25.7 <---------Smaller AICC (smaller is better) 26.4 <---------Smaller BIC (smaller is better) 25.8