Testing the Equality of Multiple Regression Curves Based on Local Polynomial Regression

迴歸曲線相似性檢定方法

洪偉喬(Wei-qiao Hong)

指導教授 ： 黃禮珊

國立清華大學/理學院/統計學研究所/碩士(2013年)

https://doi.org/10.6843/NTHU.2013.00631

摘要

In our study we extend the classical Analysis of Covariance (ANCOVA) for testing parallelism of lines between groups to nonparametric ANCOVA for testing the equality or parallelism of multiple regression curves. Our approach is based on local polynomial regression and extends the nonparametric F-tests in Huang and Chen (2008). We derive explicit expressions for the nonparametric ANCOVA table and develop nonparametric F-tests for testing the equality or parallelism of multiple curves based on assumptions of a common range of design points, homoscedastic Gaussian errors, and the same bandwidth and kernel function for fitting the curves. Simulation results indicate that the new approach is comparable to existing procedures.

關鍵字

Analysis of Covariance ； Local polynomial regression

並列摘要

In our study we extend the classical Analysis of Covariance (ANCOVA) for testing parallelism of lines between groups to nonparametric ANCOVA for testing the equality or parallelism of multiple regression curves. Our approach is based on local polynomial regression and extends the nonparametric F-tests in Huang and Chen (2008). We derive explicit expressions for the nonparametric ANCOVA table and develop nonparametric F-tests for testing the equality or parallelism of multiple curves based on assumptions of a common range of design points, homoscedastic Gaussian errors, and the same bandwidth and kernel function for fitting the curves. Simulation results indicate that the new approach is comparable to existing procedures.

參考文獻

equality of several functions: Analysis of single-unit ring-rate curves across multiple

[2] BOWMAN, A. W. and AZZALINI, A. (1997). Applied Smoothing Techniques for Data

[3] BOWMAN, A. W. and YOUNG, S. (1996). Graphical comparison of nonparametric

[4] CHAO, C. P. (2012). A simulation study of bandwidth selection for local regression

using full-information criteria. Master thesis, Nation Tsing Hua University, Hsinchu,