Title:
CALCULATION TECHNIQUE FOR ELIMINATING "MULTICOLLINEARITY" OR THE LIKE IN REGRESSION ANALYSIS, AND OBTAINING PARTIAL REGRESSION COEFFICIENT INDICATING CONTRIBUTION TO PROPER OBJECTIVE VARIABLE OF EXPLANATORY VARIABLE, AS MANAGEMENT MATERIAL
Document Type and Number:
Japanese Patent JP2021068456
Kind Code:
A
Abstract:
To solve the problem that a partial regression coefficient bi of an explanatory variable Xi of a cause indicating contribution to an effect Y is alienated from theory and experience, due to the presence of a large number of numerical values with a fairly high degree of correlation to an objective variable Y among selected explanatory variables of causes, in multiple regression analysis that is applied for relationship analysis between an industrial effect Y and its causes X.SOLUTION: Provided is a calculation technique for eliminating "multicollinearity" or the like. In order to eliminate this discrepancy, a determinant of correlation coefficients of an explanatory variable Xi and an effect Y of a normal equation in multiple regression analysis is created, and by using a diagonal element Xi that is a main force variable of a matrix since numeric balance of both right and left sides is not achieved, a portion overlapping with Xi is removed from Xj of a subordinate non-diagonal factor, and then, the remainder is used as a new non-diagonal factor of the matrix in a normal equation for which bi of a new correlation determinant is obtained.
Inventors:
SHIRAI MASAHIRO
Application Number:
JP2020183948A
Publication Date:
April 30, 2021
Filing Date:
October 15, 2020
Export Citation:
Assignee:
SHIRAI MASAHIRO
International Classes:
G06F17/16; G06F17/15; G06Q30/02