On estimating the optimal process parameters in quality engineering using generalized linear models approach

Issue title: Statistical Estimations in Complex Problems

Guest editors: Alexander Topchishvili

Article type: Research Article

Authors: Das, Rabindra Nath

Affiliations: Department of Statistics, The University of Burdwan, Burdwan, West Bengal, India. E-mail: rabin.bwn@gmail.com | Minneapolis, USA

Abstract: Generally, response surface designs are used (in quality improvement experiments) in estimating the optimal level combinations of the process parameters. In an industrial process, the most important problem is to predict the operating condition that optimize a response of interest, and simultaneously minimizes the process variability. In modern quality engineering, dual response surface approach was introduced to achieve this goal. Some researchers have proposed to use the generalized linear models to derive the joint mean and variance models, instead of separate mean and variance models as in the dual response surface approach. This article illustrates (based on two real examples) how the generalized linear models approach can be used to achieve the goal. The present analyses and interpretations (related to these two examples) are completely different from all the earlier research findings.

Keywords: Dual response surface approach, joint generalized linear models, multiplicative model, non-constant coefficient of variation, structured dispersion

DOI: 10.3233/MAS-140293

Journal: Model Assisted Statistics and Applications, vol. 9, no. 3, pp. 201-211, 2014