Nonparametric Predictive Methods for Bootstrap and Test Reproducibility

摘要: This thesis investigates a new bootstrap method, this method is called Nonparametric Predictive Inference Bootstrap (NPI-B). Nonparametric predictive inference (NPI) frequentist statistics approach that makes few assumptions, enabled by using lower and upper probabilities to quantify uncertainty, explicitly focuses on future observations. In the NPI-B we use sample of n observations to create + 1 intervals draw one value uniformly from interval. Then this added data process repeated, now with n+1 observations. Repetition leads sample, which therefore not taken actual but consists values in whole range possible observations, also going beyond sample. We explore NPI-B for on finite intervals, real line non negative observations, compare it other methods via simulation studies show NPI-B method works well as prediction method. The NPI presented for reproducibility probability (RP) some nonparametric tests. Recently, there has been substantial interest reproducibility probability, where not only its estimation definition and interpretation are uniquely determined classical statistics framework. The nature provides natural formulation of inferences RP. It used derive bounds RP (known as NPI-RP method) if consider large sizes, computation of these difficult. predict values (they NPI-B-RP values) some nonparametric Reproducibility of tests an important characteristic practical relevance test outcomes.