Kimura et al.: Kalman filtering the delay-difference equation 



687 



squares methods give similar results if there is only 

 measurement error. Bias in parameter estimates for 

 both methods tend to be small in this case. The Kalman 

 filter was superior to nonlinear least squares, as mea- 



sured by RMSE, if there is only process error. Both 

 methods tended to give biomass estimates that were 

 positively biased, and nonlinear least-squares estimates 

 were dramatically skewed to the right. A similar result 



