This paper is concerned with an anisotropy-based estimation problem for a time-varying sensor network with an information theoretic description of statistical uncertainty in the external disturbance. The latter is assumed to be a random sequence with a bounded anisotropy level. The network model involves a number of sensors with a specified communication scheme. The sensors are considered to be non-ideal, and packet dropouts of sensors appear with given probabilities. The estimation problem under consideration is reduced to a partially convex optimization problem. The designed estimation protocol secures boundedness of the anisotropic norm for the system which relates the estimation error to the input. The results are applied to a quadcopter altitude estimation problem as an example.