响应变量删失情况下线性模型的FIC模型选择和模型平均

本文给出了响应变量随机右删失情况下线性模型的FIC (focused information criterion) 模型选择方法和光滑FIC 模型平均估计方法, 证明了兴趣参数的FIC 模型选择估计和光滑FIC 模型平均估计的渐近正态性, 通过随机模拟研究了估计的有限样本性质, 模拟结果显示, 从均方误差和一定置信水平置信区间的经验覆盖概率看, 兴趣参数的光滑FIC 模型平均估计均优于FIC, AIC (Akaikeinformation criterion) 和BIC (Bayesian information citerion) 等模型选择估计; 而FIC 模型选择估计与AIC 和BIC 等模型选择估计相比, 也表现出了一定的优越性. 通过分析原发性胆汁性肝硬化数据集, 说明了本文方法在实际问题中的应用.

FIC model selection and model averaging for linear model with censored response

In this paper, we consider the FIC model selection and model averaging approach for the censored linear model. We show the asymptotic normality of the proposed FIC model selection and smoothed FIC model averaging estimator of the focused parameter. We carry a simulation study to investigate the finite property of the proposed estimator. It results out that the proposed estimator is superior to some traditional model selection methods, like AIC and BIC, in light of MSE of the estimator and the empirical coverage probability of confidence interval. We illustrated the proposed method via an analysis of a primary biliary cirrhosis (BPC) data.