Experimental designs for system assessment and improvement when noise factors are correlated

Taguchi´s strategy for tolerance design provides insight into the transmission of variability from noise factors to the system response. Under assumptions of independent noise factors, simple orthogonal designs and analysis allow for straightforward assessment of system sensitivity and evaluation of alternative system configurations. However, in certain cases the assumptions of independent noise factors are invalid. How, then, should an experiment be designed to allow similar sensitivity analysis? We illustrate our results for a model of a simple circuit, where internal noise factors may be positively correlated due to environmental conditions.