K | loglik | λopt | AICbm |
---|
1.00 | -1593.02 | 0.35 | 3192.04 |
2.00 | -1402.21 | 0.35 | 2814.41 |
3.00 | -1310.29 | 0.45 | 2634.57 |
4.00 | -1230.15 | 0.40 | 2478.30 |
5.00 | -1164.85 | 0.40 | 2351.70 |
6.00 | -1142.70 | 0.35 | 2311.39 |
7.00 | -1127.57 | 0.35 | 2285.14 |
8.00 | -1107.63 | 0.45 | 2249.26 |
9.00 | -1102.10 | 0.45 | 2242.19 |
10.00 | -1096.30 | 0.45 | 2234.60 |
- The table presents the optimal values of the transformation parameter λ that were obtained from fitting Model (2) with the boxcoxmix package [21] using the analysis data set. In addition, the respective log-likelihood and AICbm values (evaluated at the optimal λ values) are shown for varying numbers of mass points K. Following [21], the information criterion was defined as AICbm=−2 log(L)+2·(p+2K). Hence the AIC values in the fourth column cannot be directly compared to the AIC values presented in Fig. 2