An Application of Nonlinear Regression in Research and Development: A Case Study from the Electronics Industry
Karen Kafadar · Technometrics · 1994
Industry poses many challenges to the statistician, one of which is the formulation of the problem so that a technical solution can be applied. A vector demodulator, an instrument familiar to communications engineers, provides a vector representation of an input signal by representing its (possibly time-varying) amplitude and phase into two orthogonal components. To ensure the accuracy of the decomposition required for precise signal transmission, calibration adjustments to the instrument must be calculated and applied. The mathematical formulation of this problem is described and related to complex demodulation, a technique familiar to statisticians, and the statistical solution to the calibration problem using nonlinear regression is proposed. Implementation is a key component to the complete solution; during this phase, many more immediate and important issues arise. This case study illustrates the importance of not only deriving the technical solution to a statistical problem but also of following through to the implementation and production stage.