##Statistics Seminar##\\ Department of Mathematics and Statistics
^ **DATE:**|Tuesday, March 10, 2026 |
^ **TIME:**|12:15pm -- 1:15pm |
^ **LOCATION:**|WH 100E|
^ **SPEAKER:**|Zifan Huang, Binghamton University|
^ **TITLE:**|From roots to strict zero crossings: developing a well-posed definition of the Buckley-James estimator|
**Abstract**
The Buckley–James estimator (BJE) is a classical extension of least squares for linear regression with right-censored responses. Because the Buckley–James estimating function is nonsmooth and piecewise linear, defining the BJE simply as a root of the estimating equation can be ill-posed: roots may fail to exist, may be non-unique, or may form flat regions that include uninformative candidates. This talk develops a well-posed definition of the BJE through a sequence of examples, each motivating a refinement from roots to zero crossings (ZC), then to strict zero crossings (SZC), and finally to a refined multivariate SZC based on sign changes and the geometry of connected zero-crossing components.