Empirical Likelihood for A Heteroscedastic Partial Linear Model
Guo-Liang Fan,Han-Ying Liang,Hong-Xia Xu
DOI: https://doi.org/10.1080/03610921003597229
2011-01-01
Abstract:Consider the heteroscedastic regression model Y-i = X-i(tau)beta + g(T-i) + sigma(i)e(i), 1 <= i <= n, where sigma(2)(i) = f(Z(i)), beta is a p x 1 column vector of unknown parameter, (X-i, T-i, Z(i)) are random design points, Y-i are the response variables, g(.) is an unknown function defined on the closed interval [0, 1], {e(i), F-i} is a sequence of martingale differences. When f is known and unknown cases, we propose the empirical log-likelihood ratio statistics for the parameter beta. For each case, a nonparametric version of Wilks' theorem is derived. The results are then used to construct confidence regions of the parameter. Simulation study shows that the empirical likelihood method performs better than a normal approximation-based approach.