Abstract
Robust parameter design attempts to make the design or process insensitive to noise by appropriately choosing levels for the controllable factors. The robustness is achieved by exploiting the controllable-by-noise interactions. However, there exist noise factors which cannot be controlled as levels in experiments. In this article, ageneral methodology is developed for analyzing the controllable-by-noise interactions using linear regression models when noise factors are observed as covariates. This procedure is performed by backward or forward selection of variables in the linear regression model including dummy variables.
