SHI Peide;LI Guoying
Journal of Systems Science and Complexity. 1994, 7(1): 67-077.
This paper is concerned with the L1-norm estimators for the partly linear modelwhere (T_1,X_1, Y_1),…(T_n, X_n, Y_n) are independent random(d +2)-vectors such that K_i is real-valued, X_i is a d-vector of explanatory variables, and T_i is another explanatory variable ranging over a nondegenerate compact interval; u_i is arandom error; β_0 is a d-vector of parameters; and g_0(.) is an unknown function, which is m(≥0) times continuously differentiable and its mth derivative satisfies a Holder condition with exponent γ∈(0, 1]. A piecewise polynomial is used to approximate go(.). The considered estimators of β0 and g_0(t) are respectively and satisfyingwhere is a class of piecewise polynomials of degree m. Under some mild conditions, it is shown that the underlying estimators attain the convergence rate where being a constant in Condition A4.