Multiple-input multiple-output (MIMO) radars operate by simultaneously transmitting multiple independent waveforms. This facilitates improved angle-estimation performance by enabling the use of sparse antenna arrays without the ambiguities that occur when sparse arrays are used in conventional radars. Angle-estimation performance can be characterized in terms of the local error-performance bound given by the Cramér-Rao bound and in terms of the threshold point given by the SNR at which the estimator deviates significantly from the Cramér-Rao bound. In this paper, we extend results of Bliss, Forsythe, and Richmond on angle-estimation performance as a function of transmit waveform covariance for a MIMO radar. The analysis described in the above work is dependent upon an estimate or test location of a target position. Here, we provide a framework for a Bayesian extension that incorporates knowledge of the priors on the target position probability density. This information affects both the Cramér-Rao bound and the threshold SNR. Consequently, it affects waveform and system optimization.