The problem of sequential fixed-precision estimation of the minimum point of a quadratic regression function is investigated. Stochastic approximation methods are mentioned. One presents also some sequential fixed-precision procedures for the minimum of a random variable, i.e. for the lower bound of the support of its distribution function. One attempts to apply these procedures to estimation of the minimal value of a regression function. Asymptotically consistent fixed-precision estimation is considered.
W tej części opisane będą wyniki Profesora Ryszarda Zielińskiego w dziedzinie klasycznej aproksymacji stochastycznej, analizy powierzchni odpowiedzi oraz konstrukcji zbiorów ufności o zadanej precyzji.
EN
Professor Ryszard Zieliński's results in stochastic approximation, extremal experimental design in the framework of response surface analysis and fixed-precision set estimation are outlined. First, he proposed a randomized version of Fabian's (1967) gradient estimate in the Kiefer-Wolfowitz procedure, which reduced the number of required observations and improved the rate of convergence. Second, when considering response surface analysis and experimental designs for the gradient estimation, he constructed a randomized simplex design which resulted in the unbiased estimator. Third, he gave a method to construct confidence sets with prescribed accuracy (i. e. the width and the confidence level) by sampling independent copies of a process of interest. Professor Ryszard Zieliński's results in stochastic approximation, extremal experimental design in the framework of response surface analysis and fixed precision set estimation are outlined. First, he proposed a randomized version of Fabian's (1967) gradient estimate in the Kiefer-Wolfowitz procedure, which reduced the number of required observations and improved the rate of convergence. Second, when considering response surface analysis and experimental designs for the gradient estimation, he constructed a randomized simplex design which resulted in the unbiased estimator. Third, he gave a method to construct confidence sets with prescribedaccuracy (i. e. the width and the confidence level) by sampling independent copiesof a process of interest.
4
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Two concepts of optimality corresponding to Bayesian robust analysis are considered: conditional Γ-minimaxity and stability. Conditions for coincidence of optimal decisions of both kinds are stated.
5
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
A homogeneous Poisson process (N(t),t ≥ 0) with the intensity function m(t)=θ is observed on the interval [0,T]. The problem consists in estimating θ with balancing the LINEX loss due to an error of estimation and the cost of sampling which depends linearly on T. The optimal T is given when the prior distribution of θ is not uniquely specified.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.