A flexible class of time-varying link-based continuous-time random network models
based on counting processes is considered, and a rigorous analysis of corresponding
maximum likelihood estimators is presented. The model parameters are allowed to
be functions of time. The presented asymptotic analysis assumes the time horizon
tending to infinity. Our model allows for both Markovian and non-Markovian structures.
We present results on asymptotic normality of corresponding local maximum likelihood
estimators, and illustrate the finite sample performance of the estimation procedure
through numerical studies. The talk reports on joint work with Alexander Kreiss,
Heidelberg and Wolfgang Polonik, Davis.