Bayesian methods have been widely applied in vibroacoustics, architectural acoustics and signal processing in acoustics. Bayesian framework is capable of not only parameter estimation, optimization (lower level of inference), but also model selection (higher level of inference). These two levels of inference have already substantiated their wide range of applicability in acoustics. Furthermore, Bayesian probabilistic analysis is naturally useful for quantification of uncertainties. This work focuses on analyzing sound energy decays often encountered in experimentally measured data. Several advanced methods, such as nonlinear regressions, Bayesian methods, and artificial neural networks have been developed to cope with energy decay analysis being challenging in vibro-acoustics and air-borne sound in enclosures. Using these methods, a wide range of data resolutions can meet the need of energy decay analysis. Yet for high efficiency, lower resolutions may be preferable, still adequately representing energy decay processes. This paper discusses conditions of representing energy decay processes by desirable, sufficiently less data points for higher efficiency of data analysis. At the same time, increased efficiency inevitably brings uncertainties. Using Bayesian decay analysis in foreground, the analysis uncertainties are investigated against those of experimental measurements. This paper quantifies uncertainties for leveraging adequate accuracies and the analysis efficiency.