Saddlepoint Approximations with Applications

Modern statistical methods use complex, sophisticated models that can lead to intractable computations. Saddlepoint approximations can be the answer. Written from the user’s point of view, this book explains in clear language how such approximate probability computations are made, taking readers from the very beginnings to current applications. The core material is presented in chapters 1-6 at an elementary mathematical level. Chapters 7-9 then give a highly readable account of higher-order asymptotic inference. Later chapters address areas where saddlepoint methods have had substantial impact: multivariate testing, stochastic systems and applied probability, bootstrap implementation in the transform domain, and Bayesian computation and inference. No previous background in the area is required. Data examples from real applications demonstrate the practical value of the methods. Ideal for graduate students and researchers in statistics, biostatistics, electrical engineering, econometrics, and applied mathematics, this is both an entry-level text and a valuable reference.

Preface; 1. Fundamental approximations; 2. Properties and derivatives; 3. Multivariate densities; 4. Conditional densities and distribution functions; 5. Exponential families and tilted distributions; 6. Further exponential family examples and theory; 7. Probability computation with p*; 8. Probabilities with r*-type approximations; 9. Nuisance parameters; 10. Sequential saddlepoint applications; 11. Applications to multivariate testing; 12. Ratios and roots of estimating equations; 13. First passage and time to event distributions; 14. Bootstrapping in the transform domain; 15. Bayesian applications; 16. Non-normal bases; References; Index.

Explains in simple language how saddlepoint approximations make computations of probabilities tractible for complex models. No previous background in the area is required as the book introduces the subject from the very beginning. Many real data examples show the methods at work. For statisticians, biostatisticians, electrical engineers, econometricians, and applied mathematicians.

Ronald W. Butler is Professor of Statistics at Colorado State University. His statistical work on prediction, likelihood methods, and saddlepoint methods is widely known. His more recent work concerns the study of complex stochastic systems, reliability, and survival analysis, with applications to electrical engineering and medical statistics. In applied mathematics, he has made important contributions to the approximation of hypergeometric functions with matrix and vector arguments.