关于AR(p)模型参数的LAD估计收敛性

ON THE CONVERGENCE OF LAD ESTIMATES OF AR(p)MODEL CONSTANTS

一、引言和定理的叙述由于在实际问题中,具有无限方差的误差可能适合许多现象(参看[1]—[4]),所以估计下述自回归模型(1.1)的回归参数有重大的意义.

The least absolute deviation estimates L(N),from N deta points,of the auto-regressive constants a=(a_1,…, a_p)~2 for a stationary autoregressive model are shownto have the property that N~δ‖L(N)-a‖converges to zero on prabability for δ<1/α where the disturbances are i.i.d.attracted to stable law of index α,0<α<1.