In this paper we focus our attention on the estimation of dynamic discrete systems in which the observed signal is not always related to all components in the state vector, but on the contrary, it can be considered incomplete, since we do not know if a particular observed signal contains information regarding all components or not. We develop a very simple framework to deal with these situations by introducing Bernoulli random variables in the observation equation. The algorithm derived can be considered as an extension of Nahi [8] algorithm for uncertain observations. Mathematics Subject Classification: 62M20, 60G35