Here I examine how to determine the sensitivity of the LIGO, VIRGO, and LAGOS gravitational wave detectors to sources of gravitational radiation by considering the process by which data are analyzed in a noisy detector. By constructing the probability that the detector output is consistent with the presence of a signal, I show how to (1) quantify the uncertainty that the output contains a signal and is not simply noise, and (2) construct the probability distribution that the signal parameterization has a certain value. From the distribution and its mode I determine volumes $V(P)$ in parameter space such that actual signal parameters are in $V(P)$ with probability $P$. If we are {\em designing} a detector, or determining the suitability of an existing detector for observing a new source, then we don't have detector output to analyze but are interested in the ``most likely'' response of the detector to a signal. I exploit the techniques just described to determine the ``most likely'' volumes $V(P)$ for detector output corresponding to the source. Finally, as an example, I apply these techniques to anticipate the sensitivity of the LIGO and LAGOS detectors to the gravitational radiation from a perturbed Kerr black hole.

37 pages (plus 6 figures), LaTeX/REVTEX