考虑了删失分位数变系数回归模型的~FIC 准则, 并基于~FIC 准则给出了兴趣参数的模型选择和平均估计. 为了全面反映响应变量的分布信息, 克服异常值和重尾模型误差, 文章对响应变量的不同分位数水平进行建模, 因此与普通最小二乘方法相比更为稳健. 在较为一般的条件下, 证明了所提估计的渐近性质, 通过模拟实验研究了估计的有限样本性质, 用所提方法分析了手机用户的游戏时间数据.

We derive the focused information criterion for the censored quantile varying coefficient regression model to run model selection and model averaging for some focused parameters. We model different levels of quantiles of the censored response to depict the comprehensive characteristics of the distribution of the response and deal with outliers and heavy tailed model error, which makes the proposed method robuster than the ordinary least square procedure. Under general conditions, we show the large sample property of the proposed estimator. We conduct simulations to study the finite sample property of the estimator and apply the proposed method to analyze the dataset of the smart phone to model the playing time of the users of a game application in China.