Abstract
This paper presents the detailed development of a linear model for the computed solution to the linear system Ax = b; this model consists of an initial transcribing error plus a random part to model round-off. Discussed are general implications and techniques for the appli-cation of the model towards diminishing the effects of both.