Abstract: Abnormal values of an objective variable are removed. A degree of association between an objective variable and a plurality of explanatory variables is calculated. A plurality of explanatory variables having a high degree of association are extracted. A degree of independence between the explanatory variables is calculated. A plurality of candidates of explanatory variables with a high possibility of having a great effect on the objective variable are selected based on the degree of association and the degree of independence. An explanatory variable having a high rate of contribution to the objective variable is selected from among the candidates, based on the cumulative contribution rate, and a regression equation is calculated to estimate a value of the objective variable. The same processing is repeated, using the difference as a new objective variable and explanatory variables except the explanatory variable used to obtain the difference as new explanatory variables.

Abstract: Abnormal values of an objective variable are removed. A degree of association between an objective variable and a plurality of explanatory variables is calculated. A plurality of explanatory variables having a high degree of association are extracted. A degree of independence between the explanatory variables is calculated. A plurality of candidates of explanatory variables with a high possibility of having a great effect on the objective variable are selected based on the degree of association and the degree of independence. An explanatory variable having a high rate of contribution to the objective variable is selected from among the candidates, based on the cumulative contribution rate, and a regression equation is calculated to estimate a value of the objective variable. The same processing is repeated, using the difference as a new objective variable and explanatory variables except the explanatory variable used to obtain the difference as new explanatory variables.

Abstract: A systematic yield Ysi is obtained from a yield Yi, Wafer data through are sorted in numeric order of a monitored quantity X, the sorted data is classified into m groups each having approximately equal number of data, the central value is obtained for each of X and Ys in the respective wafer groups, a correlation coefficient between X and Ys is calculated for the central values, and if the coefficient is greater than a predetermined value, a regression equation of Ys with respect to X is determined. The processes of the steps 30 through 34 are performed for a number of monitored quantities. In a case where the coefficient of the correlation between X and Ys is more than a predetermined value, this X is selected and a multiple regression equation of Ys with respect to only the selected ones is determined.