摘要
CONSIDER the nonparametric median regression model Y<sub>ni</sub> = g (x<sub>ni</sub>) + ε<sub>ni</sub>, 1≤i≤n （1）where g: [0, 1]|→R is an unknown function of interest; {x<sub>ni</sub>, 1≤i≤n} is the fixed design pointsin the interval [0, 1]; {ε<sub>ni</sub>, 1≤i≤n} is a triangular array of row iid random variables having medi-an zero, and {Y<sub>ni</sub>, 1≤i≤n} is the observation. For each x∈[0, 1], n≥1, let D<sub>nj</sub>（x） denotethe jth nearest neighber of x, j=1, 2,…, n, i. e. {D<sub>n1</sub>（x）, D<sub>n2</sub>（x）,…, D<sub>nn</sub>（x）} is a trans-position of {x<sub>n1</sub>, x<sub>n2</sub>,…, x<sub>nn</sub>} satisfying |D<sub>n1</sub>（x）-x|≤|D<sub>n2</sub>（x）-x|≤…≤|D<sub>nn</sub>（x）-x|.The tie is broken by the chronological order. The
