Inverse solution for parameter estimation of computer simulation by an empirical Bayesian code tuning method

Figures & data

Table 1. Summary of c* estimates by various methods and prior moments.

Method	(2 × SE)	(2 × SE)	(2 × SE)	(2 × SE)	Function value
PMLE	0.626 (0.578)	3.313 (2.920)	1.826 (1.357)	0.673 (0.567)	−1150.96
Empirical Bayesian ()	0.510 (0.370)	0.0	2.780 (0.802)	1.638 (0.633	−1147.86
Empirical Bayesian ()	0.486 (0.537)	1.042 (2.086)	2.794 (0.959)	1.581 (0.993)	−1147.96
Prior mean	1.828	1.649	1.828	1.502
2 × Prior SD	5.598	4.322	5.598	1.072
Note: The two times standard deviations (SDs) of c* estimates are given in parentheses.

Figure 1. Plots of residual vs. predicted y values for the empirical Bayesian estimation method: (a) ASDEX tokamak and (b) PDX tokamak. Octagon: computer data, and cross: experimental data.

Figure 2. Confidence contours for and based on the empirical Bayesian estimation method. Nominals are approximately 99%, 95% and 90% with the centre at and .