本文關(guān)鍵詞：雙刪失數(shù)據(jù)下共享脆弱性模型半?yún)?shù)有效估計(jì)　出處：《中國(guó)科學(xué)技術(shù)大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 脆弱性模型 半?yún)?shù)效用 EM算法 蒙特卡洛積分

【摘要】：在這篇文章中,我們針對(duì)左右刪失的生存數(shù)據(jù),也稱雙刪失數(shù)據(jù),研究其生成的脆弱性模型。這個(gè)模型對(duì)當(dāng)前生存模型的領(lǐng)域有一定的延伸,具體來(lái)說(shuō),我們把右刪失數(shù)據(jù)在脆弱性模型當(dāng)中的應(yīng)用擴(kuò)充到具有額外左刪失數(shù)據(jù)的更復(fù)雜的情況。在這個(gè)模型中,我們采用似然函數(shù)方法來(lái)估計(jì)未知參數(shù),在估參過程中采用核心算法是EM算法�；谖覀兘⒌哪Ｐ吞厥庑浴渲泻袩o(wú)限維參數(shù),在研究中借助非參數(shù)最大似然估計(jì)法(NPMLE)估計(jì)無(wú)限維參數(shù);此外對(duì)于其他參數(shù),由于沒有顯式形式,估計(jì)會(huì)運(yùn)用到牛頓迭代法。在估參的過程中會(huì)涉及到計(jì)算量過大的問題,所以我們會(huì)對(duì)EM算法進(jìn)行改善,采用MCEM算法來(lái)降低計(jì)算量。在有限維參數(shù)中,運(yùn)用半?yún)?shù)最大似然方法后要建立相應(yīng)的漸近特性。之后文章后面會(huì)討論運(yùn)用自助法估計(jì)的標(biāo)準(zhǔn)差相容性問題。于此同時(shí),我們查找一些數(shù)據(jù)擬合到建立的模型中運(yùn)用新算法評(píng)價(jià)其估計(jì)量和穩(wěn)健性。
[Abstract]:In this paper, we study the generated vulnerability model for the left and right censored survival data, also known as double-censored data. This model has a certain extension of the current survival model domain, specifically. We extend the application of right censored data in the vulnerability model to more complex cases with additional left censored data. In this model, we use the likelihood function method to estimate unknown parameters. In the process of parameter estimation, the core algorithm is EM algorithm, based on the particularity of our model, which contains infinite dimensional parameters. In this study, the nonparametric maximum likelihood estimation (NPMLEA) is used to estimate the infinite dimensional parameters. In addition, for other parameters, because there is no explicit form, the estimation will be applied to Newton iteration method. In the process of estimating parameters, the calculation will be too large, so we will improve the EM algorithm. The MCEM algorithm is used to reduce the computation cost in finite dimensional parameters. The asymptotic properties of the semi-parametric maximum likelihood method should be established. The compatibility of the standard deviation estimated by the self-help method will be discussed later in this paper. At the same time. We find some data fitting to the established model and use the new algorithm to evaluate its estimator and robustness.
【學(xué)位授予單位】：中國(guó)科學(xué)技術(shù)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：O212.1

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 羅季;;Monte Carlo EM加速算法[J];應(yīng)用概率統(tǒng)計(jì);2008年03期

，

本文編號(hào)：1385058