报告时间： 2022-11-09 10:00——2022-11-09 11:00

报告地点： 腾讯会议： 426-781-414

报告题目： Efficient auxiliary information synthesis for semiparametric mixture cure model

报告人：王晓光

主办单位： 数据科学与人工智能学院 数理统计研究中心

【专家简介】

王晓光，大连理工大学数学科学学院副教授，博士生导师，本硕博毕业于吉林大学数学学院。主要从事大数据统计分析、生存分析和半参数模型统计推断等方面的研究工作。已在Biometrics, Statistics in Medicine等统计学主流期刊上发表研究论文40余篇，出版教材2部。担任中国现场统计学会第十、十一届理事，中国数学会概率统计学会第十一届理事。主持国家自然科学基金面上项目和青年项目等。担任大连理工大学数学科学学院金融与统计研究所党支部书记、副所长。

【讲座摘要】

Exploiting aggregated auxiliary information from external studies is becoming increasing common to improve the statistical analysis of a modestly sized internal individual level data. In this paper, we propose an auxiliary information synthesis estimation procedure to utilize subgroup survival information at multiple time points under the semiparametric mixture cure model by adopting a variance reduction technique named control variates, together with the development of a J-test statistic to check the necessary homogeneity assumption. Specifically, the auxiliary information is summarized via estimating equations and then converted into data-dependent quantities which possess the characteristic of converging to zero in probability. To adaptively accommodate population heterogeneity, this approach is further extended to a penalized auxiliary information synthesis methodology. The key idea is to introduce several nuisance parameters among which nonzero ones suggest a violation of the homogeneity assumption, followed by constructing a unified minimization problem, of which the initial approach is just a special case by treating all new parameters as zero. Moreover, these methodologies can be adjusted when the uncertainty is not negligible. Under the assumption of homogeneity and heterogeneity respectively, we establish the asymptotic properties of the proposed approaches from the perspective of efficiency gain, consistency, asymptotic normality and oracle property. Based on simulations with various settings, we demonstrate the good practical performances. We illustrate the proposed methods with an invasive breast cancer survival study.

撰稿：孙晓霞   审核：富宇     单位：数据科学与人工智能学院