| 摘要 |
PURPOSE: To create an objective acceptance criterion by means of a staistical and analytical processing operation and to reflect erroneous judgment data to an acceptance criteion creation process by a method wherein a category which sorts an object to be inspected is decided and many pieces of sample data which are to belong to every category are instructed. CONSTITUTION: When an object which is composed of two categories (e.g., a good product and a defective product) is to be discriminated, the average, the variance, the covariance and the correlation coefficient of both categories are computed on the basis of sample data, and the distance from the center of gravity of both categories of a sample which is obtained newly is computed by using them. In a discrimination analysis, sample code data groups 33, 34 which have been classified and sorted are discriminated for every category, and a discrimination function z=a1×1+2a×2+c [where (z) represents a discrimination mark obtained, a1 and a2 represent coefficients, x1 and x2 represent index values and (c) represents a constant] which is used to separate all categories optimally is computed. When the variance covariance matrix of both categories is equal, the discrimination mark obtained (z) becomes the distance up to a straight line 35 from respective points.
|
