We study the parameter estimation problem over fading channels with type-based multiple access (TBMA , , ) sensor networks. The existing estimators in the literature require the knowledge of the mean of the channel coefficients which should be non-zero, and the number of alive sensors. In this paper, we propose a novel estimator for the non-zero mean fading channel when the mean of the channel coefficients and the number of alive sensors are unknown. We compute the asymptotic variance of our novel estimator and theoretically show that it is asymptotically optimal and its asymptotic performance is the same as the existing estimator. Furthermore, we study the estimation over a zero-mean Rayleigh fading channel, where the channel status information (CSI) is available at the sensor side . We study different power control schemes over Rayleigh fading channels, and propose an algorithm to optimize the parameters for the power control scheme under both average power and peak power constraints. Simulation results corroborate our analysis.