抄録
The reliability of MPN as a point estimate of the MPN without exclusion to the ordinary two dilution schemes is of interest. Three different sources of the error have been, for years, discussed and treated for obtaining reliable MPN values: improbability of the output quanta, truncation of quantal level to three or so, and positive bias in the MPN value. Thus, first, the two intrinsically different ways of estimating improbability are critically reviewed, and another new way proposed by Kohno and Fukunaga is introduced. Second, the truncation of quantal level to three or so is discussed. Consequently, the use of a program available for computing MPN as well as estimating the improbability is claimed. Thirdly, two different ways of treating the bias in the MPN are introduced and criticized. In every case, the procedure is observed exactly and, if needed, examined through implementing C programs on a Unix machine.
