| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To solve the problem incapable of calculating a reasonable regression price because distortion due to "multicollinearity" exists in a calculated partial regression coefficient even though multiple regression analysis is applied to evaluation of standard residential sites of a large number of places. <P>SOLUTION: When a setting price is contrasted with standardization variables Ap and Ai (example i=a, b, c) of a price formation factor rating scale value, strong conformity is recognized in positive/negative signs and numerical values. Even though a considerable portion of a high correlation between factors i and j which causes "mujlticollinearity" does not have a direct correlation between the factors i and j, the considerable portion is occupied by an apparent correlation that uses, as a medium, Ap caused by the existence of conformity each between Ap and Ai, and between Ap and Aj. Accordingly, an apparent correlation portion of Ap medium is eliminated from Ai and Aj to suppress the numerical values of Ai and Aj, lowering a correlation coefficient rij of the factors i and j and thereby, calculating a partial regression coefficient having no distortion, which finds a reasonable regression price. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
