利用3年日数据检验尾部风险测度（基于风险调整收益）

本文链接：https://blog.csdn.net/Jarry1/article/details/106832425

此部份为经过前期的数据分析后，得到了显著结果的方法

library("readr")
library("readxl")
library("dplyr")
library(stringr)
library(PerformanceAnalytics)
setwd("e:/R/tail risk/month")
load("r.RData")
listd<-read.csv("listd.csv");listy<-read.csv("listy.csv")
r<-left_join(r,listd,by="Stkcd")
y5<-read.csv("y6.csv")%>%tbl_df()# 读取个股市场月数据
y5<-left_join(y5,listy,by="Stkcd")
sy<-read_excel("sy.xlsx")%>%tbl_df()
wy<-read_excel("wy.xlsx")%>%tbl_df()
varp<-0.05
#
a0b1=numeric();a1b1=numeric();a2b1=numeric();a3b1=numeric();a4b1=numeric()
a0b2=numeric();a1b2=numeric();a2b2=numeric();a3b2=numeric();a4b2=numeric()
a0c1=numeric();a1c1=numeric();a2c1=numeric();a3c1=numeric();a4c1=numeric()
a0c2=numeric();a1c2=numeric();a2c2=numeric();a3c2=numeric();a4c2=numeric()
a0d1=numeric();a1d1=numeric();a2d1=numeric();a3d1=numeric();a4d1=numeric()
a0d2=numeric();a1d2=numeric();a2d2=numeric();a3d2=numeric();a4d2=numeric()
a0e1=numeric();a1e1=numeric();a2e1=numeric();a3e1=numeric();a4e1=numeric()
a0e2=numeric();a1e2=numeric();a2e2=numeric();a3e2=numeric();a4e2=numeric()
a0g1=numeric();a1g1=numeric();a2g1=numeric();a3g1=numeric();a4g1=numeric()
a0g2=numeric();a1g2=numeric();a2g2=numeric();a3g2=numeric();a4g2=numeric()
a0f1=list(NA,NA,NA,NA,NA);a1f1=list(NA,NA,NA,NA,NA);a2f1=list(NA,NA,NA,NA,NA);a3f1=list(NA,NA,NA,NA,NA);a4f1=list(NA,NA,NA,NA,NA)
a0f2=list(NA,NA,NA,NA,NA);a1f2=list(NA,NA,NA,NA,NA);a2f2=list(NA,NA,NA,NA,NA);a3f2=list(NA,NA,NA,NA,NA);a4f2=list(NA,NA,NA,NA,NA)

#循环开始。。3年求尾部beta 
for(i in 0:20){
  for(h in 0:11){
    r1<-filter(r,Trdmnt>=(19950101+100*h+10000*i),Trdmnt<(19980101+100*h+10000*i))
    # 筛选出上市时期不大于5年窗口期第一个交易日的股票
    r1<-filter(r1,List<=min(r1$Trdmnt))
    k<-round(nrow(distinct(r1,Trdmnt))*varp)
    r2<-distinct(r1,Trdmnt,.keep_all = TRUE)%>%transmute(rme=Cdretwdos-Nrrdaydt/100,rme=-rme)%>%arrange(rme)
    r3<-slice(r2,(nrow(distinct(r1,Trdmnt))-k+1):n())
    r4<-transmute(r3,lrme=log(rme));a<-sum(r4)       
    α1<-1/k*a-log(as.numeric(slice(r2,nrow(distinct(r1,Trdmnt))-k))) #得出 Hill estimator
    r5<-mutate(r1,rje=-(Dretwd-Nrrdaydt/100),rme=-(Cdretwdos-Nrrdaydt/100))
    b<-group_by(r5,Stkcd)%>%arrange(rme)%>%slice(n()-round(n()*varp))%>%select(Stkcd,rme)%>%rename(frme=rme) #市场VAR
    f<-group_by(r5,Stkcd)%>%arrange(rje)%>%slice(n()-round(n()*varp))%>%select(Stkcd,rje)%>%rename(frje=rje) #个股VAR
    e<-group_by(r5,Stkcd)%>%summarise(k=round(n()*varp)) # 也可以求出k值
    # 检验分组正确性 ff<-filter(r5,Stkcd==2)%>%arrange(rje)
    # 有时用slice选择行会出错，可以用filter。
    c<-left_join(r5,f,by="Stkcd")%>%left_join(b,by="Stkcd")
    c1<-mutate(c,t=ifelse(rje>frje&rme>frme,1,0),tt=ifelse(rme>frme,1,0))
    c2<-group_by(c1,Stkcd)%>%summarise(τ=sum(t)/sum(tt)) 
    varm<-b
    varj<-f
    d<-left_join(c2,varj,by="Stkcd")%>%left_join(varm,by="Stkcd")%>%mutate(tβ=(τ^α1)*frje/frme)%>%select(-(2:4))%>%arrange(desc(tβ))%>%filter(tβ!="NA",tβ!="NaN")
    y6<-filter(y5,Trdmnt==(199801+h+100*i))
    
    a0b1[i*12+(h+1)]=mean(inner_join(y6,slice(d,1:(n()/5)),by="Stkcd")$eMretwd)
    a1b1[i*12+(h+1)]=mean(inner_join(y6,slice(d,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd)
    a2b1[i*12+(h+1)]=mean(inner_join(y6,slice(d,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd)
    a3b1[i*12+(h+1)]=mean(inner_join(y6,slice(d,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd)
    a4b1[i*12+(h+1)]=mean(inner_join(y6,slice(d,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd)
    
    a0b2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(d,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(y6,slice(d,1:(n()/5)),by="Stkcd")$Msmvosd)
    a1b2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(d,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(d,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$Msmvosd)
    a2b2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(d,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(d,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$Msmvosd)
    a3b2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(d,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(d,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$Msmvosd)
    a4b2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(d,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(d,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$Msmvosd)
    
    #  得出市场贝塔值：filter(n()>=24)%>%summarise(β=as.matrix(lm(eMretwd~ecm)$coefficients)[2]);summarise(β=cov(Mretwd,Cmretwdos)/var(Cmretwdos))
    y55<-filter(y5,Trdmnt>=(199501+h+100*i),Trdmnt<(199801+h+100*i))
    # 筛选出上市时期不大于3年窗口期第一个交易日的股票
    y55<-filter(y55,List<=min(y55$Trdmnt))
    dd<-group_by(y55,Stkcd)%>%filter(n()>=24)%>%summarise(β=as.matrix(lm(eMretwd~ecm)$coefficients)[2])%>%filter(β!="NA")
    g<-left_join(d,dd,by="Stkcd")%>%mutate(cβ=tβ-β)%>%select(1,4)%>%arrange(desc(cβ))%>%filter(cβ!="NA")
    a0c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$eMretwd)
    a1c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd)
    a2c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd)
    a3c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd)
    a4c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd)
    # 作差：a5c1[i*12+(h+1)]=mean(inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$eMretwd)-mean(inner_join(y6,slice(g,(n()/5*4+1):n()),by="Stkcd")$eMretwd)
    
    a0c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$Msmvosd)
    a1c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$Msmvosd)
    a2c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$Msmvosd)
    a3c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$Msmvosd)
    a4c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd,inner_join(y6,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$Msmvosd)
    # 作差： a5c2[i*12+(h+1)]=weighted.mean(inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(y6,slice(g,1:(n()/5)),by="Stkcd")$Msmvosd)-weighted.mean(inner_join(y6,slice(g,(n()/5*4+1):n()),by="Stkcd")$eMretwd,inner_join(y6,slice(g,(n()/5*4+1):n()),by="Stkcd")$Msmvosd)
    
    #加入capm 的收益：
    y7<-left_join(y6,dd,by="Stkcd")%>%filter(β!="NA")%>%mutate(eMretwd=eMretwd-ecm*β)
    
    a0d1[i*12+(h+1)]=mean(inner_join(y7,slice(g,1:(n()/5)),by="Stkcd")$eMretwd)
    a1d1[i*12+(h+1)]=mean(inner_join(y7,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd)
    a2d1[i*12+(h+1)]=mean(inner_join(y7,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd)
    a3d1[i*12+(h+1)]=mean(inner_join(y7,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd)
    a4d1[i*12+(h+1)]=mean(inner_join(y7,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd)
    
    a0d2[i*12+(h+1)]=weighted.mean(inner_join(y7,slice(g,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(y7,slice(g,1:(n()/5)),by="Stkcd")$Msmvosd)
    a1d2[i*12+(h+1)]=weighted.mean(inner_join(y7,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd,inner_join(y7,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$Msmvosd)
    a2d2[i*12+(h+1)]=weighted.mean(inner_join(y7,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd,inner_join(y7,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$Msmvosd)
    a3d2[i*12+(h+1)]=weighted.mean(inner_join(y7,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd,inner_join(y7,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$Msmvosd)
    a4d2[i*12+(h+1)]=weighted.mean(inner_join(y7,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd,inner_join(y7,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$Msmvosd)
    
    #FF3因子的结果：
    y8<-left_join(y55,sy,by="Trdmnt")
    ddd<-group_by(y8,Stkcd)%>%filter(n()>=24)%>%summarise(β=as.matrix(lm(eMretwd~mkt+smb+hml)$coefficients)[2],
                                                           SMB=as.matrix(lm(eMretwd~mkt+smb+hml)$coefficients)[3],
                                                           HML=as.matrix(lm(eMretwd~mkt+smb+hml)$coefficients)[4])
    ddd1<-filter(ddd,β!="NA",SMB!="NA",HML!="NA")
    y8<-left_join(y6,sy,by="Trdmnt")
    y9<-left_join(y8,ddd1,by="Stkcd")%>%filter(β!="NA",SMB!="NA",HML!="NA")
    y10<-mutate(y9,eMretwd=eMretwd-mkt*β-smb*SMB-hml*HML)
    
    a0e1[i*12+(h+1)]=mean(inner_join(y10,slice(g,1:(n()/5)),by="Stkcd")$eMretwd)
    a1e1[i*12+(h+1)]=mean(inner_join(y10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd)
    a2e1[i*12+(h+1)]=mean(inner_join(y10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd)
    a3e1[i*12+(h+1)]=mean(inner_join(y10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd)
    a4e1[i*12+(h+1)]=mean(inner_join(y10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd)
    
    a0e2[i*12+(h+1)]=weighted.mean(inner_join(y10,slice(g,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(y10,slice(g,1:(n()/5)),by="Stkcd")$Msmvosd)
    a1e2[i*12+(h+1)]=weighted.mean(inner_join(y10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd,inner_join(y10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$Msmvosd)
    a2e2[i*12+(h+1)]=weighted.mean(inner_join(y10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd,inner_join(y10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$Msmvosd)
    a3e2[i*12+(h+1)]=weighted.mean(inner_join(y10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd,inner_join(y10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$Msmvosd)
    a4e2[i*12+(h+1)]=weighted.mean(inner_join(y10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd,inner_join(y10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$Msmvosd)
    
    #FF5因子的结果：
    yy8<-left_join(y55,wy,by="Trdmnt")
    ddd2<-group_by(yy8,Stkcd)%>%filter(n()>=24)%>%summarise(α=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[1],
                                                             β=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[2],
                                                             SMB=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[3],
                                                             HML=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[4],
                                                             RMW=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[5],
                                                             CMA=as.matrix(lm(eMretwd~mkt+smb+hml+rmw+cma)$coefficients)[6])
    ddd3<-filter(ddd2,β!="NA",SMB!="NA",HML!="NA",RMW!="NA",CMA!="NA")
    yy8<-left_join(y6,wy,by="Trdmnt")
    yy9<-left_join(yy8,ddd3,by="Stkcd")%>%filter(β!="NA",SMB!="NA",HML!="NA",RMW!="NA",CMA!="NA")
    yy10<-mutate(yy9,eMretwd=eMretwd-mkt*β-smb*SMB-hml*HML-rmw*RMW-cma*CMA)%>%select(1:8)
    
    a0g1[i*12+(h+1)]=mean(inner_join(yy10,slice(g,1:(n()/5)),by="Stkcd")$eMretwd)
    a1g1[i*12+(h+1)]=mean(inner_join(yy10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd)
    a2g1[i*12+(h+1)]=mean(inner_join(yy10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd)
    a3g1[i*12+(h+1)]=mean(inner_join(yy10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd)
    a4g1[i*12+(h+1)]=mean(inner_join(yy10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd)
    
    a0g2[i*12+(h+1)]=weighted.mean(inner_join(yy10,slice(g,1:(n()/5)),by="Stkcd")$eMretwd,inner_join(yy10,slice(g,1:(n()/5)),by="Stkcd")$Msmvosd)
    a1g2[i*12+(h+1)]=weighted.mean(inner_join(yy10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$eMretwd,inner_join(yy10,slice(g,(n()/5*1+1):(n()/5*(1+1))),by="Stkcd")$Msmvosd)
    a2g2[i*12+(h+1)]=weighted.mean(inner_join(yy10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$eMretwd,inner_join(yy10,slice(g,(n()/5*2+1):(n()/5*(2+1))),by="Stkcd")$Msmvosd)
    a3g2[i*12+(h+1)]=weighted.mean(inner_join(yy10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$eMretwd,inner_join(yy10,slice(g,(n()/5*3+1):(n()/5*(3+1))),by="Stkcd")$Msmvosd)
    a4g2[i*12+(h+1)]=weighted.mean(inner_join(yy10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$eMretwd,inner_join(yy10,slice(g,(n()/5*4+1):(n()/5*(4+1))),by="Stkcd")$Msmvosd)
    
    ###根据月末流通市值排序大小：
    ms<-max(slice(group_by(arrange(y55,Trdmnt),Stkcd),n())$Trdmnt)
    y11<-filter(y5,Trdmnt==ms)%>%select(1,3)
    y12<-left_join(g,y11,by="Stkcd")%>%filter(Msmvosd!="NA")%>%arrange(Msmvosd)
    
    for(l in 0:4){
      a0f1[[l+1]][i*12+(h+1)]=mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*0+1):(n()/5*(0+1))),-Msmvosd),by="Stkcd")$eMretwd)
      a1f1[[l+1]][i*12+(h+1)]=mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*1+1):(n()/5*(1+1))),-Msmvosd),by="Stkcd")$eMretwd)
      a2f1[[l+1]][i*12+(h+1)]=mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*2+1):(n()/5*(2+1))),-Msmvosd),by="Stkcd")$eMretwd)
      a3f1[[l+1]][i*12+(h+1)]=mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*3+1):(n()/5*(3+1))),-Msmvosd),by="Stkcd")$eMretwd)
      a4f1[[l+1]][i*12+(h+1)]=mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*4+1):(n()/5*(4+1))),-Msmvosd),by="Stkcd")$eMretwd)
      
      a0f2[[l+1]][i*12+(h+1)]=weighted.mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*0+1):(n()/5*(0+1))),-Msmvosd),by="Stkcd")$eMretwd,
                                            inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*0+1):(n()/5*(0+1))),-Msmvosd),by="Stkcd")$Msmvosd)
      a1f2[[l+1]][i*12+(h+1)]=weighted.mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*1+1):(n()/5*(1+1))),-Msmvosd),by="Stkcd")$eMretwd,
                                            inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*1+1):(n()/5*(1+1))),-Msmvosd),by="Stkcd")$Msmvosd)
      a2f2[[l+1]][i*12+(h+1)]=weighted.mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*2+1):(n()/5*(2+1))),-Msmvosd),by="Stkcd")$eMretwd,
                                            inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*2+1):(n()/5*(2+1))),-Msmvosd),by="Stkcd")$Msmvosd)
      a3f2[[l+1]][i*12+(h+1)]=weighted.mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*3+1):(n()/5*(3+1))),-Msmvosd),by="Stkcd")$eMretwd,
                                            inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*3+1):(n()/5*(3+1))),-Msmvosd),by="Stkcd")$Msmvosd)
      a4f2[[l+1]][i*12+(h+1)]=weighted.mean(inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*4+1):(n()/5*(4+1))),-Msmvosd),by="Stkcd")$eMretwd,
                                            inner_join(yy10,select(slice(arrange(slice(y12,(n()/5*l+1):(n()/5*(l+1))),desc(cβ)),(n()/5*4+1):(n()/5*(4+1))),-Msmvosd),by="Stkcd")$Msmvosd)}}}
##循环结束。。

# 直接生成表格 
m1=numeric();m2=numeric();m3=numeric();m4=numeric();m5=numeric();m6=numeric()
for(o in 1:2){
  for(i in 1:5){
    m1[i+(o-1)*12]=mean(lapply(paste0("a",0:4,"b",o),get)[[i]])*100}
  m1[i+(o-1)*12+1]=mean(lapply(paste0("a",0:4,"b",o),get)[[1]]-lapply(paste0("a",0:4,"b",o),get)[[5]])*100
  
  for(i in 1:5){
    m1[i+(o-1)*12+6]=t.test(lapply(paste0("a",0:4,"b",o),get)[[i]])$statistic[[1]]}
  m1[i+(o-1)*12+7]=t.test(lapply(paste0("a",0:4,"b",o),get)[[1]]-lapply(paste0("a",0:4,"b",o),get)[[5]])$statistic[[1]]}

#根据差价分组
for(o in 1:2){
  for(i in 1:5){
    m2[i+(o-1)*12]=mean(lapply(paste0("a",0:4,"c",o),get)[[i]])*100}
  m2[i+(o-1)*12+1]=mean(lapply(paste0("a",0:4,"c",o),get)[[1]]-lapply(paste0("a",0:4,"c",o),get)[[5]])*100
  
  for(i in 1:5){
    m2[i+(o-1)*12+6]=t.test(lapply(paste0("a",0:4,"c",o),get)[[i]])$statistic[[1]]}
  m2[i+(o-1)*12+7]=t.test(lapply(paste0("a",0:4,"c",o),get)[[1]]-lapply(paste0("a",0:4,"c",o),get)[[5]])$statistic[[1]]}

#投资组合收益率经过CAPM模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m3[i+(o-1)*12]=mean(lapply(paste0("a",0:4,"d",o),get)[[i]])*100}
  m3[i+(o-1)*12+1]=mean(lapply(paste0("a",0:4,"d",o),get)[[1]]-lapply(paste0("a",0:4,"d",o),get)[[5]])*100
  
  for(i in 1:5){
    m3[i+(o-1)*12+6]=t.test(lapply(paste0("a",0:4,"d",o),get)[[i]])$statistic[[1]]}
  m3[i+(o-1)*12+7]=t.test(lapply(paste0("a",0:4,"d",o),get)[[1]]-lapply(paste0("a",0:4,"d",o),get)[[5]])$statistic[[1]]}

#投资组合收益率经过FF3模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m4[i+(o-1)*12]=mean(lapply(paste0("a",0:4,"e",o),get)[[i]])*100}
  m4[i+(o-1)*12+1]=mean(lapply(paste0("a",0:4,"e",o),get)[[1]]-lapply(paste0("a",0:4,"e",o),get)[[5]])*100
  
  for(i in 1:5){
    m4[i+(o-1)*12+6]=t.test(lapply(paste0("a",0:4,"e",o),get)[[i]])$statistic[[1]]}
  m4[i+(o-1)*12+7]=t.test(lapply(paste0("a",0:4,"e",o),get)[[1]]-lapply(paste0("a",0:4,"e",o),get)[[5]])$statistic[[1]]}

#投资组合收益率经过FF5模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m5[i+(o-1)*12]=mean(lapply(paste0("a",0:4,"g",o),get)[[i]])*100}
  m5[i+(o-1)*12+1]=mean(lapply(paste0("a",0:4,"g",o),get)[[1]]-lapply(paste0("a",0:4,"g",o),get)[[5]])*100
  
  for(i in 1:5){
    m5[i+(o-1)*12+6]=t.test(lapply(paste0("a",0:4,"g",o),get)[[i]])$statistic[[1]]}
  m5[i+(o-1)*12+7]=t.test(lapply(paste0("a",0:4,"g",o),get)[[1]]-lapply(paste0("a",0:4,"g",o),get)[[5]])$statistic[[1]]}

mm1<-matrix(c(m1,m2,m3,m4,m5),20,6,byrow=TRUE)
#经过规模细分的投资组合收益率经过FF5模型回归的截距项
for(o in 1:2){
  for(v in 1:5){
    for(l in 1:5){
      m5[l+(v-1)*5+(o-1)*60]=mean(lapply(paste0("a",0:4,"f",o),get)[[v]][[l]])*100}}
  for(l in 1:5){
    m5[l+(v-1)*5+(o-1)*60+5]=mean(lapply(paste0("a",0:4,"f",o),get)[[1]][[l]]-lapply(paste0("a",0:4,"f",o),get)[[5]][[l]])*100}
  
  for(v in 1:5){
    for(l in 1:5){
      m5[l+(v-1)*5+(o-1)*60+30]=t.test(lapply(paste0("a",0:4,"f",o),get)[[v]][[l]])$statistic[[1]]}}
  for(l in 1:5){
    m5[l+(v-1)*5+(o-1)*60+35]=t.test(lapply(paste0("a",0:4,"f",o),get)[[1]][[l]]-lapply(paste0("a",0:4,"f",o),get)[[5]][[l]])$statistic[[1]]}}

# 平均
for(o in 1:2){
  for(v in 1:5){
    m6[v+(o-1)*12]=mean(apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[v]]),1,mean))*100}
  m6[v+(o-1)*12+1]=mean(apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[1]]),1,mean)-
                          apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[5]]),1,mean))*100
  
  for(v in 1:5){   
    m6[v+(o-1)*12+6]=t.test(apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[v]]),1,mean))$statistic[[1]]}
  m6[v+(o-1)*12+7]=t.test(apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[1]]),1,mean)-
                            apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[5]]),1,mean))$statistic[[1]]}

mm2<-matrix(m6,4,6,byrow = TRUE)
mm3<-rbind(mm1,matrix(m5[1:30],5,6),mm2[1,],matrix(m5[31:60],5,6),mm2[2,],matrix(m5[61:90],5,6),mm2[3,],matrix(m5[91:120],5,6),mm2[4,])
write.csv(mm3,"E:/R/tail risk/table/mm.csv",row.names = FALSE)

检验市场收益率处于极端损失时的有效性

# 左右尾分布：
## 1月：
ys<-distinct(y5,Trdmnt,.keep_all = TRUE)%>%filter(Trdmnt>=199801&Trdmnt<=201812)%>%arrange(ecm)
#左尾
s<-slice(ys,1:(0.1*n()))
#右尾
s<-slice(ys,(n()-0.90*n()+1):n())
# 计算分位数值:  slice(ys,0.1*n())$ecm;slice(ys,n()-0.5*n()+1)$ecm

# 直接生成表格
##根据尾部贝塔值分组
m1=numeric();m2=numeric();m3=numeric();m4=numeric();m5=numeric();m6=numeric()
# 投资组合收益率
for(o in 1:2){
  for(i in 1:5){
    m1[i+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[i]]),s,by="Trdmnt")$zsy)*100}
  m1[i+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[1]]),s,by="Trdmnt")$zsy
                        -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[5]]),s,by="Trdmnt")$zsy)*100
  
  for(i in 1:5){
    m1[i+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[i]]),s,by="Trdmnt")$zsy)$statistic[[1]]}
  m1[i+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[1]]),s,by="Trdmnt")$zsy
                          -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"b",o),get)[[5]]),s,by="Trdmnt")$zsy)$statistic[[1]]}

#根据差价分组
for(o in 1:2){
  for(i in 1:5){
    m2[i+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[i]]),s,by="Trdmnt")$zsy)*100}
  m2[i+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[1]]),s,by="Trdmnt")$zsy
                        -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[5]]),s,by="Trdmnt")$zsy)*100
  
  for(i in 1:5){
    m2[i+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[i]]),s,by="Trdmnt")$zsy)$statistic[[1]]}
  m2[i+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[1]]),s,by="Trdmnt")$zsy
                          -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"c",o),get)[[5]]),s,by="Trdmnt")$zsy)$statistic[[1]]}

#投资组合收益率经过CAPM模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m3[i+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[i]]),s,by="Trdmnt")$zsy)*100}
  m3[i+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[1]]),s,by="Trdmnt")$zsy
                        -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[5]]),s,by="Trdmnt")$zsy)*100
  
  for(i in 1:5){
    m3[i+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[i]]),s,by="Trdmnt")$zsy)$statistic[[1]]}
  m3[i+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[1]]),s,by="Trdmnt")$zsy
                          -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"d",o),get)[[5]]),s,by="Trdmnt")$zsy)$statistic[[1]]}

#投资组合收益率经过FF3模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m4[i+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[i]]),s,by="Trdmnt")$zsy)*100}
  m4[i+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[1]]),s,by="Trdmnt")$zsy
                        -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[5]]),s,by="Trdmnt")$zsy)*100
  
  for(i in 1:5){
    m4[i+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[i]]),s,by="Trdmnt")$zsy)$statistic[[1]]}
  m4[i+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[1]]),s,by="Trdmnt")$zsy
                          -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"e",o),get)[[5]]),s,by="Trdmnt")$zsy)$statistic[[1]]}

#投资组合收益率经过FF5模型回归的截距项
for(o in 1:2){
  for(i in 1:5){
    m5[i+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[i]]),s,by="Trdmnt")$zsy)*100}
  m5[i+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[1]]),s,by="Trdmnt")$zsy
                        -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[5]]),s,by="Trdmnt")$zsy)*100
  
  for(i in 1:5){
    m5[i+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[i]]),s,by="Trdmnt")$zsy)$statistic[[1]]}
  m5[i+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[1]]),s,by="Trdmnt")$zsy
                          -semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,zsy=lapply(paste0("a",0:4,"g",o),get)[[5]]),s,by="Trdmnt")$zsy)$statistic[[1]]}

mm1<-matrix(c(m1,m2,m3,m4,m5),20,6,byrow=TRUE)
#经过规模细分的投资组合收益率经过FF5模型回归的截距项
for(o in 1:2){
  for(v in 1:5){
    for(l in 1:5){
      m5[l+(v-1)*5+(o-1)*60]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=lapply(paste0("a",0:4,"f",o),get)[[v]][[l]]),s,by="Trdmnt")$eMretwd)*100}}
  for(l in 1:5){
    m5[l+(v-1)*5+(o-1)*60+5]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=lapply(paste0("a",0:4,"f",o),get)[[1]][[l]]-lapply(paste0("a",0:4,"f",o),get)[[5]][[l]]),s,by="Trdmnt")$eMretwd)*100}
  
  for(v in 1:5){
    for(l in 1:5){
      m5[l+(v-1)*5+(o-1)*60+30]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=lapply(paste0("a",0:4,"f",o),get)[[v]][[l]]),s,by="Trdmnt")$eMretwd)$statistic[[1]]}}
  for(l in 1:5){
    m5[l+(v-1)*5+(o-1)*60+35]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=lapply(paste0("a",0:4,"f",o),get)[[1]][[l]]-lapply(paste0("a",0:4,"f",o),get)[[5]][[l]]),s,by="Trdmnt")$eMretwd)$statistic[[1]]}}

# 平均
for(o in 1:2){
  for(v in 1:5){
    m6[v+(o-1)*12]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[v]]),1,mean)),s,by="Trdmnt")$eMretwd)*100}
  m6[v+(o-1)*12+1]=mean(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[1]]),1,mean)-apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[5]]),1,mean)),s,by="Trdmnt")$eMretwd)*100
  
  for(v in 1:5){   
    m6[v+(o-1)*12+6]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[v]]),1,mean)),s,by="Trdmnt")$eMretwd)$statistic[[1]]}
  m6[v+(o-1)*12+7]=t.test(semi_join(data.frame(Trdmnt=arrange(ys,Trdmnt)$Trdmnt,eMretwd=apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[1]]),1,mean)-apply(do.call("cbind",lapply(paste0("a",0:4,"f",o),get)[[5]]),1,mean)),s,by="Trdmnt")$eMretwd)$statistic[[1]]}
mm2<-matrix(m6,4,6,byrow = TRUE)
mm3<-rbind(mm1,matrix(m5[1:30],5,6),mm2[1,],matrix(m5[31:60],5,6),mm2[2,],matrix(m5[61:90],5,6),mm2[3,],matrix(m5[91:120],5,6),mm2[4,])
write.csv(mm3,"E:/R/tail risk/table/mm.csv",row.names = FALSE)