Embedded Files
#計算VaR 利用1-1900筆資料計算第1901筆的VaR
#series VaR=NA
scalar w=1000000
scalar c_001=critical(z,0.01)
smpl 1 1900
scalar sigma=sd(Y)
scalar k2=$t2+1
series VaR[k2]=w*c_001*sigma/100
smpl full
#計算VaR 利用2-1901筆資料計算第1902筆的VaR
scalar w=1000000
scalar c_001=critical(z,0.01)
smpl 2 1901
scalar sigma=sd(Y)
scalar k2=$t2+1
series VaR[k2]=w*c_001*sigma/100
smpl full
移動視窗 (rolling window) 係指固定樣本視窗長度 (以下例子為 2年)
但樣本區間逐期移動的樣本選取方法，見以下 script 範例:
smpl 2000/01/01 2001/12/31
loop for i=1..3
smpl +1 +1
ols y const x
end loop
例:
smpl 1 1900
Full data range: 1 - 1974 (n = 1974)
Current sample: 1 - 1900 (n = 1900)
? smpl +1 +1
Full data range: 1 - 1974 (n = 1974)
Current sample: 2 - 1901 (n = 1900)
? smpl 1 500
Full data range: 1 - 1974 (n = 1974)
Current sample: 1 - 500 (n = 500)
? smpl +1 +1
Full data range: 1 - 1974 (n = 1974)
Current sample: 2 - 501 (n = 500)
? smpl 1 1900
Full data range: 1 - 1974 (n = 1974)
Current sample: 1 - 1900 (n = 1900)
#寫回圈的指令
scalar w=1000000
scalar c_001=critical(z,0.01)
smpl 1 1900
loop for j=1900..1974
scalar sigma=sd(Y)
scalar k2=$t2+1
series VaR[k2]=w*c_001*sigma/100
smpl +1 +1
end loop
#歷史模擬法計算 VaR
series VaR=NA
smpl 1 1900
loop for j=1901..1974
scalar w=1000000
scalar r=quantile(Y,0.01)
scalar k2=$t2+1
VaR[k2]=w*r/100
smpl +1 +1
end loop
移動視窗歷史模擬法的缺點
If 只移動30天則移動的天數會未包含最大的損失
Page updated
Google Sites
Report abuse