The Likelihood Surface in the Linear Functional Relationship Problem

Abstract

Summary In a recent paper, Solari has shown that no maximum likelihood solution exists in the linear functional relationship problem. When account is taken of the rounding-off errors in the observations, however, the appropriate likelihood function is bounded and a solution can be found which is approximately maximum likelihood, in the sense that the value of the likelihood at that solution is close to the global supremum. The novel form of this solution is shown and the likelihood surface in a numerical example is discussed.