Dynamically adjustable linear data discriminator

摘要

A method of discriminating between linear and nonlinear regions of measured data is based upon two hypotheses, one of which is a hypothesis that a response is merely a result of inherent randomness and the other of which is a hypothesis that the response is a result of a true nonlinear change greater than a value M. It is also based upon a relative cost between accepting the latter hypothesis when the former is true versus accepting the former hypothesis when the latter is true. A resulting probabilistic criterion (M/2)+( sigma 2/M) [1n(PH0/PH1)C] is graphically implemented for deriving offset values which can be used in a specific embodiment to construct tabular values defining linear/nonlinear regions from a presumed hypothetical straight line response.