A statistical approach implies that the commonly accepted derivation f
or the stochastic variation in line power leads to a 67% underestimate
. The 'lifetime' of a mode should be defined in terms of the reciproca
l of the linewidth, not the e-folding time of the mode. The ratio of p
ositive to negative measurement uncertainties in observed linewidth da
ta is essentially constant. Modeling the distribution of the linewidth
measurement values as an exponential produces self-consistent estimat
es of the standard deviation, and can be useful in examining error pro
pagation. The ratio of the line-power uncertainties is also nearly con
stant. However, the measurement distribution is more complex. It exhib
its behavior similar to an exponential, but has more components.