“…The approximate relative changes based on linear approximation, p(') and T(1), and those based on quadratic approximation, p(i) + (e/2)P(2) and T(1) + (e/2)T(2), analysis (LSFA) and alpha factor analysis (AFA) have a common property that an eigenvalue problem of a symmetric matrix is contained as a part in their determinating equations . Based on this property Tanaka and Odaka(1989a,b,c), Tanaka and Tarumi(1989) and have derived the influence functions for the unique and common variance matrices in PFA, MLFA, LSFA, CFA, and AFA, respectively, using the perturbation expansion of T discussed in section 5.2. Let us consider the case of PFA.…”