Computer Model Calibration or Tuning in Practice

Computer models to simulate physical phenomena are now widely available in engineering and science. Before relying on a computer model, a natural first step is often to compare its output with physical or field data, to assess whether the computer model reliably represents the real world. Field data, when available, can also be used to calibrate or tune unknown parameters in the computer model. Calibration is particularly problematic in the presence of systematic discrepancies between the computer model and field observations. We introduce a likelihood alternative to previous Bayesian methodology for estimation of calibration or tuning parameters. In an important special case, we show that maximum likelihood estimation will asymptotically find values of the calibration parameter that give an unbiased computer model, if such a model exists. However, the calibration parameters are not necessarily estimable. We also show in some settings that calibration or tuning need to take into account the end-use prediction strategy. Depending on the strategy, the choice of the parameter values may be critical or unimportant.