Semiparametric estimation of a Box-Cox transformation model with varying coefficients model

• 来源期刊: 中国科学：数学（英文版）
• 年，卷（期）: 2017,（5）
• 页码范围: 897-922
• 页数: 26页
• 语种: 英文
• DOI: 10.1007/s11425-016-0150-1

摘要

This paper considers the estimation of a Box-Cox transformation model with varying coefficient.A two-step approach is proposed in which the first step estimates the varying coefficients nonparametrically for any given parameter α in the transformation function.Then a one-dimensional search of α has been employed based on some least absolute deviation criterion function.The validity of our estimator does not require independence assumption thus is robust to the conditional heteroscedasticity.A simulation study shows a reasonably well finite sample performance.Additionally,a comprehensive empirical study has been carefully examined.