1. Monte Carlo 模擬:
特點:
隨機抽樣
獨立
計算量大
相對于historic simulation method(nonparametric method)更精確
可用于任何概率分布,任何情況
generate the data according to the desired data generating process(DGP)
Sampling error:標準誤
Sx=根號下 [Var(X)/ N]
N代表重復(fù)次數(shù)the number of replication
置信區(qū)間:[x拔 - z*Sx,x拔 + z*Sx]
short option → extreme loss → Var(x)↑ → Sx ↑
降低標準誤的方法:
① N ↑:
但是N有可能高到無法令人接受
②對偶變量法:
antithetic variates:
取補集complement,平行模擬 parallel simulation
x拔=(x1+x2)/ 2
Var(x拔)= [? var(x1)+ var(x2)+ 2Cov(x1,x2)] / 4
不用對偶變量法時:
因為x1,x2獨立,所以Cov=0
即Var(x拔)= [? var(x1)+ var(x2)] / 4
用對偶變量法時:
rou(x1,x2) < 0
Cov(x1,x2)<0
Var ’(x拔)< Var(x拔)
③控制變量法:
control variates:
control variable have a high correlation (similar to that used in? the simulation),but which properties are known prior to the simulation.
add values that are mean 0.
x* = y +(x^ - y^)
Var(x*)= Var(x^)+? Var(y^)-? 2Cov(x^,y^)
要降低抽樣誤差,即Var(x*)<? Var(x^)
即Var(y^)-? 2Cov(x^,y^)< 0
2. Boostraping 方法:重抽樣方法
特點:
generate simulated data
sampling repeatedly
和simulation一樣都充分利用了observed data
區(qū)別:
Monte Carlo:
使用observed data來估計key model parameters,比如mean、standard deviation。并且作出關(guān)于它們分布的假設(shè)an assumption of the distribution。
Bootstrapping :
直接使用observed data 來模擬出一個有相似特征的sample 。不需要模擬 observed data,也不需要作出關(guān)于它們分布的假設(shè)。
One Simple Fact:
反復(fù)抽樣的分布來源于已觀察的數(shù)據(jù)
方法:
①iid method:
observations互相獨立,無自相關(guān)性
②Circular Block Bootstrap:(CCB)
financial data 有自相關(guān)性
block size = 根號下 sample size
Limitations:
過去不能反映未來(太過依賴歷史數(shù)據(jù))
bootstrapping不能產(chǎn)生在樣本中沒有出現(xiàn)的數(shù)據(jù)
bootstrapping和simulation都會遭受“黑天鵝”(“Black Swan” problem)事件
(a good statistical model:應(yīng)該允許the probability of future losses 大于 have realized in? the past)
Ineffective Situations:
①outliers的發(fā)生和發(fā)生頻率
②bootstrapping假設(shè)數(shù)據(jù)互相是獨立的,而實際上這明顯不成立。因為數(shù)據(jù)有自相關(guān)性。
3. Random Number Generation:
類型:
①truly random number真隨機數(shù):
time consuming and difficult
②pseudorandom number偽隨機數(shù):
computer-generated random number are in fact not random at all
formula公式/算法
最簡單的類型:
從uniform distribution N(0,1)中抽取
equal chance
以初始值initial value(seed)開始
解決辦法:
初始值會影響the characteristics of generation distribution,開始影響很大,最后消失
generate more random number than required and discard previous observations
Random number reusage:
優(yōu)點:reduce the variability of the difference in? the estimates across experiments.
缺點:
①不會提高數(shù)據(jù)的精確性
②不太可能節(jié)省大量的時間
Disadvantages of simulation:
①large replications,expensive computationally
②DGP(data generating process)的不切實際的假設(shè),導(dǎo)致模擬結(jié)果less precise
③results hard to replicate
④results are experiment-specific