User Tools

Site Tools


seminars:stat:140918

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
seminars:stat:140918 [2014/09/18 14:13] nyeseminars:stat:140918 [2014/09/18 14:14] (current) nye
Line 1: Line 1:
 +<WRAP centeralign>##Statistics Seminar##\\ Department of Mathematical Sciences</WRAP>
 +
 +~~META:title =September 18, 2014~~
 +<WRAP 70% center>
 +|  **DATE:**|Thursday, September 18, 2014 |
 +|  **TIME:**|1:15pm to 2:15pm |
 +|  **PLACE:**|OW 100E |
 +|  **SPEAKER:**|Qiqing Yu (Binghamton University) |
 +|  **TITLE:**|Models allowing dependent right censoring  |
 +</WRAP>
 +\\ 
 +
 +<WRAP center box 80%>
 +<WRAP centeralign>**Abstract**</WRAP>
 +
 +The independent  right censorship (RC) model assumes that
 +the survival time  and censoring time are independent.
 +Williams  and Lagakos (W&L) (1977) propose
 +a continuous dependent RC model.
 +They  show that it
 +is equivalent to
 +a continuous independent RC model.
 +Yu //et al.//} (2012) study another dependent RC model, say Model 2,
 +without the continuity restriction, which is essentially
 + the necessary and sufficient
 +condition for the product-limit-estimator (PLE)
 +being consistent
 +for all time within the largest exact observation under the non-parametric
 +set-up.
 +However the asymptotic normality of the PLE under that model is not
 +addressed due to technical difficulties.
 +In this paper, we show that
 +(1) the continuous (but not the discrete)
 +W&L model is a special case of Model 2;
 +(2)
 +Model 2
 + is equivalent to
 +an independent RC model, extending W&L's result.
 +Thus the asymptotic normality of the PLE under Model 2
 +follows from the existing results in the literature
 +under the independent RC model.
 +
 +
 +
 +</WRAP>