On a Simple Estimation Procedure for Censored Regression Models with Known Error Distributions

L. Breiman,Y. Tsur,A. Zemel
DOI: https://doi.org/10.1214/AOS/1176349394
1993-12-01
Abstract:A simple and tractable iterative least squares estimation procedure for censored regression models with known error distributions is analyzed. It is found to be equivalent to a well-defined Huber type M-estimate. Under a regularity condition, the algorithm has excellent convergence properties. The resulting estimate is shown to be 4N-consistent and asymptotically normal. 'Statistics Department, University of California, Berkeley. Supported by NSF Grant No. DMS-8718362 2Department of Agricultural and Applied Economics, University of Minnesota 3The Jacob Blaustein Institute for Desert Research, Ben-Gurion University of the Negev, Israel A Simple Estimation Procedure for Censored Regression Models with Known Error Distribution
What problem does this paper attempt to address?