5.6Robust Estimation--Motivation--I

• See Andrews, et al. ``Robust Estimates of Location,'' Princeton University Press;

• ``Robustness in Statistics,'' edited by R. L. Laurer, G. N. Wilkinson, Academic Press (1979).

• In the maximum likelihood method we found the method that had the smallest possible Standard Deviation (asymptotically at least).

• However, as we emphasized in the study of non-parametric estimates, it is quite possible (and indeed probable) that systematic error from incorrect theoretical form of probability is large.

• In many cases, one knows reasonably well the analytic form of the ``interesting'' effect but what one observes has other totally unknown contributions.

• One must choose between parameterizing the unknown or ignoring known form of known effect.