Regression quantiles and trimmed least squares estimators for structural equations and nonlinear regression models

摘要

This thesis extends some robust estimation techniques to the structural equation model and nonlinear regression model. Bahadur representations and limiting distributions for regression quantiles and trimmed least squares estimators based on regression quantiles have been derived for both regression models. In addition, some asymptotic properties for a trimmed least squares estimator based on the quantiles of residurals of a predetermined estimator for the structural equation models has also been derived.