Abstract
The author applies the generalized Student's ratio to the analysis of the time series, considering an ordered discrete values of a time series as a vector X'. The fundamental idea is as follows: If any alternative meteorological phenomenon A (for example, the passage of a depression) does not bring about any significant effect on living bodies, a set of certain measure B of biological variation (considered as k-dimensional component vector X of X') constitutes a random sample of size N from the population B, when A occurs N times. He introduces here the generalized variance ratio H (tensor) by the method of the likelihood ratio and introduces further a new method of rejection of a set of correlated data and a new method of curve fitting, and criticizes R. A. Fisher's formal application of his method of analysis of variance on a time series and W. R. Thompson's method of rejection of a datum.
