TREND ANALYSIS - BINARY-VALUED AND POINT CASES

A sequence of occurrence times of hoods may be considered to be part o f a realization of a binary-valued time series or of a stochastic poin t process. In this paper a criterion for detecting the presence of a m onotonic trend in the rate of the process is considered. The criterion is based on linear functions of the data with the coefficients chosen to emphasize a monotonic rate. In the case that the process is statio nary and mixing, the null distribution of the test statistic is approx imately standard normal.