Statistical and algorithmic methods for fluctuation analysis with SALVADOR as an implementation.

This paper aims at removing certain long-standing impediments to more effective and widespread use of fluctuation analysis. The paper presents a method of constructing confidence intervals for mutation rates using data from fluctuation experiments. The method was inspired by a rediscovery of a little-known, not fully developed method of Lea and Coulson; substantial modifications have been made both to enhance computational efficiency and to widen the scope of the original method's applicability. A computer package named SALVADOR is presented that can be used for Monte Carlo simulation, for point and interval estimation of mutation rates, and for exploration of various hypotheses spawned by the directed mutation controversy. In addition to the maximum likelihood method, methods of considerable historical interest are also examined and included in SALVADOR to help the reader compare and assess some of the most popular methods for estimating mutation rates.