Application of Bayesian statistics and optimization to difference equation parameter estimation

The authors present a method for estimating the parameters of a difference equation from data that has random error in both the input and output measurements. A stochastic global optimization technique iteratively determines the set of parameters that maximize the posterior density function for the parameters conditioned upon the error between the measured output data and simulated data derived from the model. Examples show that errors in parameter estimates in the presence of input error can be reduced in comparison with other techniques.<
>