- 作者: 傅承德; 洪文良
- 作者服務機構: 中央研究院統計科學研究所; 國立中央大學統計學研究所
- 中文摘要: 考慮由一具有轉移機率矩陣之非齊次設限馬可夫鏈所獲得的γ個觀察值 (方程式無法摘錄) 。對於Aalenand Johansen(1978)和Phelan(1988)所提的積估計量,我們使用自助法與貝氏自助法近似 (方程式無法摘錄) ,的抽樣分佈,進而建構近似的信賴區間。本文証明在大樣本時,自助法與貝氏自助法具有第一階漸近性。有關兩者在小樣本的應用情形,本文亦做了一個模擬研究。
- 英文摘要: Many population processes in medical clinical trial, demography, epidemiology and other fieldscan be represented by censored Markov chain models. The censoring mechanism is generally useful inapplications and maintains the conditional multinomial sampling model for successive transitions. In thisarticle, we focus on setting confidence intervals for the unknown parameters of transition probabilities.That is, consider r observations (方程式無法摘錄) from a non-homogeneous censored Markovchain, with transition probability matrix For the product estimator of proposed by Aalen andJohansen (1978) and Phelan (1988), we investigate the behavior of bootstrap and Bayesian bootstrap toapproximate the sampling distribution of (方程式無法摘錄) and then construct an approximateconfidence interval. The rationale of Bayesian bootstrap is based on the conjugate beta-neutral processprior. The performance of these approximations will be justified using large sample theory as well assmall sample simulation.
- 中文關鍵字: Bayesian bootstrap; bootstrap; censored Markov chains; product estimator; transition probability matrix
- 英文關鍵字: --