The effect of noise on least-squares based estimation of nonlinear system parameters is described. While both bias and variance are affected by noise, bias is the more difficult problem to deal with. Variance can be reduced by us ing more data, but bias can, in some cases, become greater when more data is used. A parameter estimation technique, that utilizes the steady-state response of a
... [Show full abstract] fractional derivative model of the uni-directional dynamic response of a polyurethane foam block, is analyzed to illustrate why the estimation bias occurs. Also described is the influence of unmodeled dynamics on the esti mates of parameters in a model of a mode of a cantilever beam.