Sizing up the regions of unique minima in the least squares nonlinear regression

Abstract

In nonlinear regression analysis, the residual sum of squares may possess multiple local minima. This complicates finding the global minimum and adversely affects reliability of the relevant statistical methods. Identifying and sizing up the regions of a readily identifiable global minimum (RIGM) is therefore of both theo- retical and practical interest. This paper addresses the issue by using equidistant function previously introduced by the first two co-authors of this paper.