Abstract
The effect of the serial dependence of observations on the classical χ2 test is considered. For large correlations it is necessary to make the corrections to the mean and variance of χ2 given in this paper. The theory is illustrated with particular reference to the Poisson and normal Markoff processes. It is a pleasure to record my sincere thanks to Prof. M. S. Bartlett, under whose supervision thin work was done. To him I owe the suggestion of the problem, and his comments and advice were invaluable during the investigation. I am indebted to Mrs Linnert for the calculations in Tables 5·1 and 5·2.

This publication has 0 references indexed in Scilit: