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.
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 a in the transformation function. Then a one-dimensional search of a 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.
基金
supported by National Natural Science Foundation of China(Grant Nos.71171127,71471108 and 71601105)
the Open Project Program in the Key Laboratory of Mathematical Economics(SUFE)(Grant No.201309KF02)
Ministry of Education of the People’s Republic of China
the Program for Changjiang Scholars and Innovative Research Team in Shanghai University of Finance and Economics
the Innovative Research Team of Econometrics in Shanghai Academy of Social Sciences
