Chen, Min (2010) A comparison of nonlinear and nonparametric regression methods. Masters thesis, Memorial University of Newfoundland.
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
In this report, we investigate the performance of nonlinear regression and nonparametric regression with data set simulated under a nonlinear parametric model. First, we consider the nonlinear least squares estimation method for the model. Then, we apply various nonparametric regression methods such as kernel methods, spline smoothing, and wavelet version of estimators with the same model. The nonlinear least squares estimation method produces the best estimation in terms of MSE among all the regression methods. Both kernel methods and wavelet version of estimation methods produce reasonably small values of MSE. Moreover, the wavelet regression method performances best among all the nonparametric methods. The spline method produces unacceptably large MSE due to large variance of estimation. The boundary issues do exist in all the nonparametric regression methods due to less density of data points.
|Item Type:||Thesis (Masters)|
|Additional Information:||Includes bibliographical references (leaves 48-49). -- Transcript.|
|Department(s):||Science, Faculty of > Mathematics and Statistics|
|Library of Congress Subject Heading:||Regression analysis--Mathematical models--Evaluation; Smoothing (Statistics)|
Actions (login required)