欢迎光临散文网 会员登陆 & 注册

复习笔记Day113:概率论知识总结(五)

2023-03-04 22:54 作者:间宫_卓司  | 我要投稿

虽然本来不打算看第七章的,但今天早上翻了一下课本,发现有的内容我还是感兴趣的,所以选着看一些

第七章 随机序列的收敛

§7.5 依分布收敛

定义7.5.1 (1)设F_n%5Cmathbf%7BR%7D上的分布函数。称%5Cleft%5C%7B%20F_n%20%5Cright%5C%7D%20弱收敛,如果存在一个递增右连续的函数F,使得对任何F的连续点x,有%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_n%5Cleft(%20x%20%5Cright)%20%3DF%5Cleft(%20x%20%5Cright)%20,记为F_n%5Cxrightarrow%7Bw%7DF;

(2)如果%5Cxi%20_n的分布函数序列F_n弱收敛于%5Cxi的分布函数F,称%5Cxi%20_n%5CRightarrow%20%5Cxi%20,这里%5Cxi%20_n%5Cxi甚至不需要是一个概率空间上的

%5Cxi%20_n%5CRightarrow%20%5Cxi%20个人感觉是一个很弱的条件,随机序列%5Cxi_n甚至可以不收敛,例如取%5Cxi服从标准正态分布,%5Cxi_n%3D(-1)%5En%5Cxi,那么%5Cxi_n%5Cxi分布函数相同

定理7.5.1%5Cleft%5C%7B%20%5Cxi%20_n%20%5Cright%5C%7D%20%2C%5Cxi%20是随机变量,分别具有分布函数%5Cleft%5C%7B%20F_n%20%5Cright%5C%7D%20%2CF。如果%5Cxi%20_n%5Cxrightarrow%7Bp%7D%5Cxi%20,则对F的任意连续点x%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_n%5Cleft(%20x%20%5Cright)%20%3DF%5Cleft(%20x%20%5Cright)%20

式中的%5Cxrightarrow%7Bp%7D指的是依概率收敛,即%5Cmathbb%7BP%7D%20%5Cleft(%20%5Cleft%7C%20%5Cxi%20_n-%5Cxi%20%5Cright%7C%5Cge%20%5Cvarepsilon%20%5Cright)%20%2C%5Cforall%20%5Cvarepsilon%20%3E0

定理7.5.2 (%5Ctext%7BSkorohod%7D)设F_n%2Cn%5Cge1F%5Ctextbf%7BR%7D上的分布函数,如果F_n%5Cxrightarrow%7Bw%7DF,则存在概率空间%5Cleft(%20%5COmega%20%2C%5Cmathscr%7BF%7D%20%2C%5Cmathbb%7BP%7D%20%5Cright)%20与其上的随机变量%5Cleft%5C%7B%20%5Cxi%20_n%20%5Cright%5C%7D%20%2C%5Cxi%20使得

(1)%5Cxi_n点点收敛于%5Cxi

(2)F_nF分别是%5Cxi_n%5Cxi的分布函数

这个定理的证明思路上和定理 5.1.2的证明有点像,先证明了在连续点处,F_%7Bn%7D%5E%7B-1%7D点点收敛于F,然后再记%5Cxi%20_n%3DF_%7Bn%7D%5E%7B-1%7D%5Cleft(%20%5Ceta%20%5Cright)%201_%7B%5Cleft%5C%7B%20%5Ceta%20%5Cnotin%20D%20%5Cright%5C%7D%7D%2C%5Cxi%20%3DF%5E%7B-1%7D%5Cleft(%20%5Ceta%20%5Cright)%201_%7B%5Cleft%5C%7B%20%5Ceta%20%5Cnotin%20D%20%5Cright%5C%7D%7D,其中%5Ceta服从%5B0%2C1%5D上的均匀分布,1_%7B%5Cleft%5C%7B%20%5Ceta%20%5Cnotin%20D%20%5Cright%5C%7D%7D1_%7B%5Cleft%5C%7B%20x%5Cnotin%20D%20%5Cright%5C%7D%7D%5Cleft(%20%5Ceta%20%5Cright)%20的意思,DF的不连续点集。那么%5Cxi_n逐点收敛于%5Cxi,且因为D至多可列,所以%5Cxi_n%2C%5Cxi的分布函数就是F_n%2CF

定理7.5.3 (%5Ctext%7BHelly-Bray%7D)设有分布函数F_n%2CF。则F_n%5Cxrightarrow%7Bw%7DF当且仅当对任何有界连续函数f%5Cint_%7B%7D%5E%7B%7D%7Bf%5Cmathrm%7Bd%7DF_n%7D%5Crightarrow%20%5Cint_%7B%7D%5E%7B%7D%7Bf%5Cmathrm%7Bd%7DF%7D

定理7.5.4 任何分布函数列%5Cleft%5C%7B%20F_n%20%5Cright%5C%7D%20有一个子列%5Cleft%5C%7B%20F_%7Bk_n%7D%20%5Cright%5C%7D%20弱收敛

这个定理的证明看的我有点懵,我按照自己的理解写一下好了

%5Cmathbf%7BR%7D中的一个稠密点集D%3D%5C%7Bx_n%3An%5Cge1%5C%7D,那么对于任何给定好的n,可以找到点列%5Cleft%5C%7B%20F_n%5Cleft(%20x_1%20%5Cright)%20%5Cright%5C%7D%20收敛子列%5Cleft%5C%7B%20F_%7Bk_%7Bn%7D%5E%7B%5Cleft(%201%20%5Cright)%7D%7D%5Cleft(%20x_1%20%5Cright)%20%5Cright%5C%7D%20,接下来,对于点列%5Cleft%5C%7B%20F_%7Bk_%7Bn%7D%5E%7B%5Cleft(%201%20%5Cright)%7D%7D%5Cleft(%20x_2%20%5Cright)%20%5Cright%5C%7D%20,又可以找到收敛子列%5Cleft%5C%7B%20F_%7Bk_%7Bn%7D%5E%7B%5Cleft(%202%20%5Cright)%7D%7D%5Cleft(%20x_2%20%5Cright)%20%5Cright%5C%7D%20,一直这样做下去,就可以找到函数序列%5C%7BF_n%5C%7D的一个收敛子列%5Cleft%5C%7B%20F_%7Bk_%7Bn%7D%5E%7B%5Cleft(%20n%20%5Cright)%7D%7D%5Cleft(%20x%20%5Cright)%20%5Cright%5C%7D%20,使得其满足%5Cleft%5C%7B%20F_%7Bk_%7Bn%7D%5E%7B%5Cleft(%20n%20%5Cright)%7D%7D%5Cleft(%20x_i%20%5Cright)%20%2Ci%5Cle%20n%5Cright%5C%7D%20收敛,所以记k_%7Bn%7D%5E%7B%5Cleft(%20n%20%5Cright)%7D%3Dk_n的话,

就有%5Cleft%5C%7B%20%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_%7Bk_n%7D%5Cleft(%20x_i%20%5Cright)%20%2Cx_i%5Cin%20D%20%5Cright%5C%7D%20收敛

接下来去证明%5Cleft%5C%7B%20F_%7Bk_n%7D%20%5Cright%5C%7D%20确实是弱收敛的,首先记%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_%7Bk_n%7D%5Cleft(%20x_i%20%5Cright)%20%3Dy_i,那么如果弱收敛到的函数就被%5C%7By_i%5C%7D所唯一确定了,所以不妨定义

F%5Cleft(%20x%20%5Cright)%20%3D%5Cbegin%7Bcases%7D%0A%09y_i%2Cx_i%5Cin%20D%5C%5C%0A%09%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7Dy_%7Bx_0%5Cleft(%20n%20%5Cright)%7D%2Cx%3Dx_0%5Cnotin%20D%5C%5C%0A%5Cend%7Bcases%7D

其中%5C%7Bx_0(n)%5C%7D%5Csubset%20Dx_0%5Cleft(%20n%20%5Cright)%20%5Cdownarrow%20x_0,下面来证明确实有%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_%7Bk_n%7D%5Cleft(%20x_0%20%5Cright)%20%3DF%5Cleft(%20x_0%20%5Cright)%20,其中x_0是连续点,那么此时F_%7Bk_n%7D%5Cleft(%20x_%7Bi_m%7D%20%5Cright)%20%5Cle%20F_%7Bk_n%7D%5Cleft(%20x_0%20%5Cright)%20%5Cle%20F_%7Bk_n%7D%5Cleft(%20x_%7Bj_n%7D%20%5Cright)%20,其中x_%7Bn_j%7D%2Cx_%7Bn_i%7D%5Csubset%20D%2Ci%2Cj%3D1%2C2%2C%5Ccdots%20%2Cn,令n%5Crightarrow%20%5Cinfty%20,可得y_%7Bn_j%7D%5Cle%20%5Cunderset%7Bn%5Crightarrow%20%5Cinfty%7D%7B%5Clim%7DF_%7Bk_n%7D%5Cleft(%20x_0%20%5Cright)%20%5Cle%20y_%7Bn_i%7D,这里严格来说要取上下极限,但是我懒得打了。再取x_%7Bi_m%7D%5Cuparrow%20x_0%2Cx_%7Bj_n%7D%5Cdownarrow%20x_0,并分别令i%2Cj趋于无穷,这就证明了结论

(这个证明的后半段我竟然想了一个晚上,wssb)

图片好像快100张了,接下来的两章只能下一篇再发布了···

           



复习笔记Day113:概率论知识总结(五)的评论 (共 条)

分享到微博请遵守国家法律