1. The Power Law:幂定律
高峰肥尾(eg:t分布)
V:the value of variable
P(V>X)= K * X^(-α)
(K,α > 0且为常数)
X↑ P↓
2. Correlation and Covariance:
Covn = E(xnyn)- E(xn)E(yn)= E(xnyn)
E(xn)=E(yn)=0
rou(linear) = Cov(xn,yn)/ 6xn * 6yn
3. Update Covariance:
①EWMA:
Covn = 入Cov n-1 +(1-入)xn-1 yn-1
②GARCH(1,1):
Covn = YCovL + α xn-1 yn-1 + βCov n-1
4. Alternative measures of correlation:
即可衡量线性,也可衡量非线性:
Spearman’s rank correlation 和 Kendal’s tao
rou between -1 and 1
returns independent:rou = 0
relationship increase:rou > 0
relationship decrease:rou < 0
①Spearman’s rank correlation:
ordinal
nonparametric approach
没有 linear correlation estimator有效
可作为additional robustness check
性质:
①不受outliers值影响
②X,Y单调性不变,rou值不变
②Kendall’s tao:
ordinal
nonparametric approach
性质:
①不受outliers值影响
②X,Y单调性不变,tao值不变
正序对Concordant:Xi>Xj,Yi>Yj 或者
Xi<Xj,Yi<Yj
异序对Discordant:Xi>Xj,Yi<Yj 或者
Xi<Xj,Yi>Yj
既不是正序对,也不是异序对:Xi=Xj,Yi=Yj
网友评论