m*****n 发帖数: 3575 | 1 基本模型
dS/S = k(a-logS)dt + sdW
变换
Z=logS
dZ= k(a- ss/2k - Z)dt + sdW
Y=exp(kt)Z 消去dt前面的Z得到additive brownian motion:
dY= k(a-ss/2k)*exp(kt)dt + s*exp(kt)dW
得Z(t)的解析解:
Z=Y/exp(kt)
~ N{ Z0exp(-kt)+(a-ss/2k)[1-exp(-kt)], var= ss[1-exp(-2kt)]/2k }
如果给出一系列数据
S1.......Sn
自然也已知
Z1.......Zn
请问怎么用模型拟合数据解出
k a 和 s
特别是 k 和 s
怎么求?
急问,多谢,40伪币+ 解释得越详细越好 | w**********y 发帖数: 1691 | 2 several different methods, based on Maximum likelihood, least square error,
or method of moment.
a short paper for reference here:
http://www.investmentscience.com/Content/howtoArticles/MLE_for_OR_mean_reverting.pdf | m*****n 发帖数: 3575 | |
|