This is a problem I face in my work. I don't have a answer yet. All discussions are welcome. We are observing a time series C(t), the n th observation is realized at time t = tn, with value Cn. C(t) is monotonic increasing in t. Theoratical, we know C(t) / t converges (say, to R, an unknown number). We would like to have a good rule for stopping time, T, so that C(T)/T is a reasonable approximation of R. www.ddhw.com
|