The joint decentralized estimation and control problem of a random access (RA) communication network is considered. It is shown that application of the certainty equivalence principle results in a nonlinear separation rule for the control. If the minimum-mean-squared-error (mmse) estimate of the channel backlog is used to control the retransmission probability, an unstable RA broadcast channel can be stabilized for input traffic rates not exceeding e-1. The performance of the mmse estimator-controller is also analyzed via simulation.