We present a method that accommodates missing data in longitudinal datasets of the type usually encountered in economic and social applications. The technique uses various extensions of missing at random' assumptions that we customize for dynamic models. Our method, applicable to longitudinal...