Super-short time series’ parameters estimate on the base of granular data about record intervals between episodes

An improved approach to behavior rate estimates on the base of data about minimum, maximum and usual interval between behavior episodes is considered. The mathematical model of this behavior is a Poisson stochastic process; the observations are respondents' natural language answers about mentioned intervals. The considered method takes account of data granularity; it is based on the analysis of ordinal statistics and method of randomization. In the paper, there are several examples of the estimates applications to some other types of socially significant behavior, including risky behavior.