M-estimation in nonlinear regression for longitudinal data

The longitudinal regression model Z(i)(j) = m(theta(0), X-i(T-i(j))) + epsilon(j)(i), where Z(i)(j) is the jth measurement of the ith subject at random time T-i(j), m is the regression function, Xi(T-i(j).) is a predictable covariate process observed at time T-i(j). and epsilon(j)(i) is a noise, is studied in marked point process framework. In this paper we introduce the assumptions which guarantee the consistency and asymptotic normality of smooth M-estimator of unknown parameter theta(0).