两样本秩次统计量的几乎处处界

THE ALMOST SURE BOUND OF TWO-SAMPIE RANK ORDER STATISTICS

  • 摘要: 设 X1,…,Xn i.i.d.X1~F,Y1,…,Yn,i.i.d.Y1~G,这里 F 和 G 是两个一维连续分布函数.以 Ri 记 Xi 在合并样本(X1,…,Xm,Y1,…,Yn)中的秩,且设φ(μ)定义于(0,1),φN(n)定义于1/(N+1),…,N/(N+1).本文给出了如下结果:在φ(x)与φN(x)满足一定条件下\beginaligned & \left|s_N-\mu_N\right| \leqslant A \sqrt\frac\log \log NN . . \quad \text a.s \\ & \left|S_*^*-\mu_N\right| \leqslant A \sqrt\frac\log \log NN . \quad \text a.s \endaligned其中\begingatheredS_N=\frac1m \sum_i=1 \phi\left(R_i /+1\right) \\ S_N^*=\frac1m \sum_i=1^m \phi_N\left(R_i /\right) . \\ \mu_N=\int_-\infty^\infty \phi(H(x)) \cdot d F(x) \\ H(x)=\lambda_N F(x)+\left(1-\lambda_N\right) G(x), \lambda_N=\fracmN,\endgathered。

     

    Abstract: Under appropriate conditions,we give the almost sure bound of two-sample rank order statistic and prove it analogous to the partral sums of independent identically distributed random variables.

     

/

返回文章
返回