Calibration of computer models with heteroscedastic measurement errors
Computer models are commonly used to represent a wide range of real systems, but they often involve some unknown parameters. Estimating the parameters by collecting physical data becomes essential in many scientific fields, ranging from engineering to biology. However, most of the existing methods are developed under a homoscedastic error assumption. Motivated by an experiment of plant relative growth rates where replicates are available, we propose a new calibration method for the physical data with heteroscedastic measurement errors. Numerical examples demonstrate that the proposed method not only yields accurate parameter estimation, but it also provides accurate predictions for physical data when the measurement errors are heteroscedastic. We compare this new method to the standard approach used to interpret growth data and give additional examples of how this approach can be used to produce more statistically robust conclusions from computer models of biology and biochemistry in general. We additionally outline how the approach can be used to determine optimal sampling locations.
READ FULL TEXT