Your model is very interesting. Can I ask some questions? Please ignore them if the answers can reveal some intrinsic features of your system.
1) Do you use linear functions f(t,i), or more complicated polynomials are allowed?
2) What is the average distance between t(i) and t(i+1)? 1 day? 1 week? 1 month? How strongly this distance is allowed to fluctuate for a given system?
3) Do you use least-squares method to approximate C(t) by f(t,i)? Do you choose t(i) and the polynomial coefficients of f(t,i) which minimize the norm of the difference C(t) — f(t,i)?
4) Do you consider N(t) to be a Gaussian noise?
Thank you. I shall try to estimate, how reliable the «predictions» of this approach could be. Probably this was already done in some book.
