Abstract: The revue of results obtained by the author this year in the area of disturbance of statistical stability of physical quantities and processes is presented. Two new sensitive parameters characterizing their statistical instability in finite observation interval are proposed. For known and new parameters of statistical instability unit measures that give possibility to characterize the level of disturbance are introduced. It is shown by modeling that STD corridors of unit measures weakly depend from the type of the distribution. It is confirmed that important role in disturbance of statistical stability plays specific fluctuation of expectation of the process that generates changes of average expectation. Dependence of parameters of statistical instability from normalized sample variance of expectation average is found. It is shown by modeling that positive correlation calls increasing of statistical instability and negative correlation – decreasing of it. Under any type correlation increasing of sample size calls increasing of statistical stability and the process trends to statistically stable state. It is found that changes of variance influence to statistical stability of the process. Under some relations of parameters unpredictable component of the process fades-out and under other ones – fades-in. It is demonstrated high sensitivity of new parameter of statistical instability. It is shown that fluctuations of maximum day temperature and minimum day temperature in two cities (Moscow and Kiev) are instable and fluctuation of day precipitation is near stable. Disturbance of statistical stability of temperature begins from some observation weeks.

Keywords: statistical instability, theory of hyper-random phenomena, uncertainty, probability.

ACM Classification Keywords: G.3 Probability and Statistics

Link:

DISTURBANCE OF STATISTICAL STABILITY (PART II)

Igor Gorban

http://foibg.com/ijita/vol18/ijita18-4-p02.pdf