仔细看了三篇adaptive 文献
7楼:A Markov Chain Model for the Adaptive CUSUM....
10楼:An Adaptive CUSUM Procedure for Signaling Process Variance
24楼:An adaptive EWMA control chart....
所选用的weight function,score function都不相同
就以10楼 WACUSUM_S2而言
m=1.2
h=max(m,(l+0.5)*q);
aa=log(h)/(1-1/h);
a1=(k-i-0.5)*w/(1-1/h)+aa;
a2=(k-i+0.5)*w/(1-1/h)+aa;
b1=(l-(1-p)*(j+0.5))*q/p;
b2=(l+1-(1-p)*(j+0.5))*q/p;
适用于WACUSUM_S2,然不适用于WACUSUM_ln(S2)
因为几乎exp(temp)之后都是1,
造成 T2-T1=0,所以ARL非常小.
adaptive应该不能如此转换.
建议你想办法拿到最基础的这篇文献
Brook and Evans (1972),An Approach to the Probability Distribution of the CUSUM Run Length
再来参考修改
另这篇文献也试试
Shu,L.,Jiang,W.and Wu,Z.,Adaptive CUSUM Procedures with Markovian Mean Estimation 2008