1986 年 52 巻 484 号 p. 3198-3206
In this paper, investigating the error included in experimental data statistically, the theoretically proper weighting function used in curve fits on the frequency domain is made clear. Many kinds of curve fits have been proposed and have come into wide use in the stage of industrial design. However, there are still some ambigous points even in the theory of the curve fits. For example, it is not clear what should be used as the weighting function in the curve fits on the frequency domain. Up to now, the reciprocal of the magnitude of the transfer function has been used in order to fit the skirts of the transfer function well, generally. This is not right from the viewpoint of statistical theory. Therefore, it is explained theoretically and positively that the variance of the experimental data should be used as the weighting function.