References 283 swap) did not always lead to performance breakdown. Therefore, subspace swap is a necessary, but not sufficient, condition for DOA estimation breakdown, with other factors also influencing outlier production, such as statistical variations of the source power and manner of distribution of the signal subspace power across the sample noise subspace. These investigations once again point out the need for performance assess- ment tools other than the standard large-sample asymptotics when dealing with finite-sample threshold region behavior. The power of the RMT/GSA asymptotic approach is particularly evident here, as is the capability of expected likeli- hood to provide nonclairvoyant solution assessment in scenarios outside the ML-breakdown region but within SNR and/or sample support regimes whether or not other estimation methods fail. In summary, the finite-sample performance assessment objectives laid out at the beginning of the chapter can be achieved using new applications of like- lihood ratio formulations as well as Random Matrix theory, providing new tools for the assessment of DOA estimation algorithms in the low training- sample environment, which is increasingly dominating the field of adaptive processing. REFERENCES