Burr Type XII分配逐步型型一區間設限資料之貝氏分析－使用R2WinBUGS套件

Bayesian Analysis of Progressive Type I Interval Censored Data from Burr Type XII Distribution - Using R2WinBUGS Package

林余昭(Yu-Jau Lin)

《智慧科技與應用統計學報》 12卷2期 (2014/12) Pp. 75-86

摘要

最近，貝氏統計分析已廣泛應用在各領域，但是它在實務上的困難是使用者常須自己寫程式來執行統計計算，而WinBUGS軟體以簡單的語法即可有效執行馬可夫鍊蒙地卡羅法來估計模型參數，大量簡化了貝氏統計計算的困難度。本研究以R軟體的R2WinBUGS套件來對Burr type XII分配下逐步型型一區間設限資料作貝氏統計分析。我們首先在R的命令環境下作資料處理，再利用R2WinBUGS套件呼叫WinBUGS統計軟體以執行統計計算，而所得的結果也回傳R作進一步分析。我們用大規模的模擬研究來驗證貝氏分析的正確性，再將方法應用到112位骨髓腫瘤患者存活時間的真實資料。

關鍵字

設限資料 ； 馬可夫鍊蒙地卡羅法

並列摘要

We first consider the Bayesian estimation of progressive type I interval censored data from Burr type XII distribution. The R2WinBUGS package, that calls WinBUGS from R, provides an easy way for researchers to operate data manipulation in R and do parameter estimation by applying Markov chain Monte Carlo method in WinBUGS. The progressive type I interval censored data from Burr type XII distribution are firstly generated in R. With the use of R2WinBUGS package, R calls WinBUGS in the batch mode to execute the Gibbs sampling scheme that applies the Metropolis-Hastings algorithm by specifying the posterior likelihood, proportional to the product of the likelihood and parameter prior distribution. The results are again returned to R for further statistical analysis. Extensive simulation studies are conducted to investigate the performance of the developed methods. Finally, a real data set containing 112 patients with plasma cell myeloma is analyzed for illustration.

並列關鍵字

R ； WinBUGS ； Censoring ； Markov chain Monte Carlo