Inferring lifetime distributions from kinetics by maximizing entropy using a bootstrapped model.

A bootstrapped model is used to improve the lifetime distribution recovered using the maximum entropy method from kinetics that involves overlapping exponential and distributed phases. The model defaulted to in the limit of low signal-to-noise is iteratively derived from the data to counter the tendency of regularization methods to over-smooth sharp features while under-smoothing broad ones. Upon each revision, some of the lifetime distribution is focused and the rest is blurred. This differential blurring can produce distributions that are virtually free of artifacts. The change in the result obtained upon a reasonable change in the default model provides a useful measure of the uncertainty in the lifetime distribution. In particular, the widths of peaks may not be well determined.