Signal processing forum
A.P. Tomaniac · IEEE Signal Processing Magazine · 1998
Asymptotic analyses are abundant in the literature. However, some people (practitioners or not) question their usefulness. Others argue that they are useful. We do the same here. More exactly, we try to explain briefly why "asymptomania" is not a "dangerous disorder", but rather an addiction to a useful tool. One can readily identify the following possible goals of an asymptotic (in N or SNR, etc.) analysis [where N is the number of available data samples, and SNR is the signal-to-noise ratio]: (a) derivation (or estimation) of the theoretical performance of a parameter estimator or detector; (b) analytical comparisons of the detection or estimation performance of a number of methods; and (c) the optimal choice of the user's variables of a detection or estimation method.