A type of biased estimators for linear models with uniformly biased data

Abstract.

The objective of this paper is the comparison of various types of estimators that can be used in linear models with uniformly biased data. This particular case refers to adjustment problems where the available measurements are affected by a common, unknown and uniform offset. The classic least-squares (LS) unbiased estimators for this type of models are reviewed in detail, and some additional remarks on their properties and performance are given. Furthermore, a family of biased estimators for linear models with uniformly biased data is introduced, which has the potential to provide better performance (in terms of mean squared estimation error) than the ordinary LS unbiased solutions. A number of different regularization viewpoints that can be equivalently associated with these biased estimators are presented, along with a discussion on various selection strategies that can be employed for the choice of the regularization parameter that enters into the biased estimation algorithm.