'; } else{ document.getElementById(PLTS[cur1].portletId).innerHTML=response; } } catch(e){} var regx=new RegExp(']*?>([\\s\\S]*?)<\\/scr'+'ipt\\s*>|]*?\\/>','img'); var arr; while((arr = regx.exec(response)) != null){ if(arr.length){ if(arr.length >=2){ var regx2 = new RegExp("]*?\\bsrc=\\s*('[^>]*'|\"[^>]*\"|[\\S]*)\\s*\\/?>",'im'); var srcArr=regx2.exec(arr[0]); if(srcArr!=null){ var src= srcArr[1]; if(src){ if((src.charAt(0)=='"') || (src.charAt(0)=="'")){ src=src.substr(1,src.length-2); } loadScript(src); } } else{ execJavaScript(arr[1]); } } } } } cur1++; loadps(); } } function loadScript(s){ try{ var sc=document.createElement('script'); sc.type='text/javascript'; sc.language='javascript'; sc.src=s; document.body.appendChild(sc); }catch(e){} } function execJavaScript(s){ try{ var sc=document.createElement('script'); sc.type='text/javascript'; sc.language='javascript'; sc.text=s; document.body.appendChild(sc); }catch(e){} } window.onload=loadps;


主  题: Cox Regression with Nonignorable Survival-Dependent Missing Covariate Values

内容简介: Analysis with censored survival time in clinical and epidemiological studies often encounters missing covariate data and a missing at random assumption is commonly adopted, which assumes that missingness depends on observed censored data, the minimum of survival and censoring time. Although missingness is likely related with time of survival, sometimes it is not reasonable to assume that censoring affects missingness of a covariate, especially when covariates are measured at baseline. If missingness of a covariate depends on survival time (and other covariates with no missing values), but not censoring, then missingness is nonignorable since survival time may be censored, and data analysis is challenging. In this article we propose a method in Cox regression with survival-dependent missing covariates, which is shown to produce consistent and asymptotically normal estimators of parameters. Our method is based on inverse propensity weighting with both censored and non-censored survivals. The propensity depending on non-censored survival is estimated nonparametrically by product kernel regression. The
finite-sample performance of the proposed estimators is examined through simulation and by an application to a real-data example.

报告人: 邵军      教授    博导    国家“千人计划”创新人才

时  间: 2018-01-04    10:30

地  点: 竞慧东楼302

举办单位: 理学院  统计科学与大数据研究院