Parameter design, which is popularized by japanese quality expert Taguchi, is to design products and manufacturing processes such that performance characteristics are robust to uncontrollable noise effects. The idea of parameter design is intrinsically different form such traditional quality control techniques as screening, control charts, inspection, etc. in that it is implemented at the stage of design and development. A series of quality control activities using parameter design is called quality engineering. His stated objective of parameter design is to find the conditions of design parameters that minimize average quality loss or maximize a quantity called signal-to-noise(SN) ratio. However, he gives no explanations for the connection between these two measures and a poor explanation about generic solution procedures. In this thesis, a model and an analysis method for parameter design is presented considering a linear relation between the input signal and the ideal value of a performance characteristic. This study presents two criteria for problem classification and a new performance measuere, expected quality loss after adjustment, which is proved to be equivalent to Taguchi's SN ratio approximately. On the basis of these, a two-step optimization procedure is proposed for each problem classified. Further, by establishing the scaling invariant property, it is shown that parameter design can make a contribution toward the improvement of flexibility in manufacturing process.