Robust Parameter Estimator for Non-Linear Growth Curves.
NAVAL POSTGRADUATE SCHOOL MONTEREY CALIF
Pagination or Media Count:
The parameter estimation methods considered in this thesis are the weighted Least-Squares and Weighted Huber for some non-linear growth models. The properties of these parameter estimators derived from simulated data by means of 1 weighted and unweighted least-squares and 2 weighted and unweighted Huber robust estimation are compared. The error components of the simulated data are long-tailed and non-normal. The performance of mis-specified models is considered. Author
- Statistics and Probability