Title:  Partial Regression

Abstract:

	Partial regression refers to the situation of
	fitting regression models when the "good" data
	are mixed not only with outliers but also with
	a separate unknown regression process.  A few
	authors have discussed fitting mixtures of
	regressions, but assume models are known for
	each.  Here, we describe how a partial mixture
	estimator, L2E, can be extended to partial
	regression, providing a technique which is also
	resistant to outliers.