Some interval and trend statistics with non-Gaussian initial data distribution
2018; American Institute of Physics; Linguagem: Inglês
10.1063/1.5064895
ISSN1935-0465
AutoresVolodymyr Myrhorod, I. Hvozdeva, В. В. Демиров,
Tópico(s)Gaussian Processes and Bayesian Inference
ResumoAn approach to the research of interval and trend statistics of cumulative sums, Hald-Abbe and Fisher (F – criterion) with a non-Gaussian distribution of the initial data is proposed and justified. Estimates of the shifts at significance levels and the values of the confidence probabilities of these statistics are determined depending on the kurtosis of the initial data a priori distribution. The a priori statistical model of data generation adopted in the researches was chosen as a model for expanding the distribution of the initial data into the Edgeworth series.
Referência(s)