An improved saddlepoint approximation based on the negative binomial distribution for the general birth process (Q1855614)

This paper proposes an improved saddlepoint approximation which is only slightly more complex than \textit{H.E. Daniels}' approximation [J. Appl. Probab. 19, 20-28 (1982; Zbl 0482.60080)] but has certain advantages. The assumption that the count distribution of a birth process follows a negative binomial distribution, when the birth rates are linear increasing, enables the authors to derive the waiting time distribution for any birth rate sequence which is a permutation of a linear increasing sequence. This distribution then serves as a leading term for saddlepoint approximations for a general birth rate process. The obtained results are exact not only for negative binomial distributions, but also in Poisson (constant birth rates) and binomial (linear decreasing rates) cases as well as in any case where the birth rates can be ordered in a linear sequence. Numerical comparisons also demonstrate that the proposed saddlepoint approximation is often more accurate than Daniels' one.