Identification of a hereditary system with distributed delay (Q1061085)

The maximum likelihood method for identifying the parameter function in a linear hereditary system is considered. The delay is allowed to be distributed and so the classical theory is not directly applicable, since even if the maximum likelihood method is valid the estimates obtained may not be consistent. The system model is \[ dX_ t=\int^{0}_{-b}\alpha (\theta)X_{t+\theta}d\theta dt+dW_ t \] and \(\alpha\) is estimated by using finite-dimensional approximations in the space \(L^ 2[-b,0]\). This gives rise to the finite-dimensional maximum likelihood estimate \({\hat \alpha}{}^ T(\theta)\) for the process \(\{X_ t\), \(t_ 0\leq t\leq T\}\). Sufficient conditions are given for the consistency of the estimate, i.e. \[ \lim_{T\to \infty}\int^{T}_{t_ 0}| {\hat \alpha}^{(T)}(\theta)-\alpha (\theta)|^ 2d\theta =0, \] in probability.