期刊名称：Brazilian Journal of Probability and Statistics
印刷版ISSN：0103-0752
出版年度：2002
卷号：16
期号：1
页码：25-38
出版社：Brazilian Statistical Association
摘要：In several regression problems monotonicity is a key feature of the underlying re-gression function, although in some cases the observations are not strictly mono-tonic due to random error. There are some other cases in which the observationsare monotonic by nature. In all such cases the fitted curve should possess thismonotonicity in order to explain the dataset. In the present paper, the dataseton the development of the world record on men's 100 m sports is considered foranalysis. Using Bayesian methodology the fittings of the data is described bytwo methods, namely using monotone spline and the lo cal regression techniqueof O'Hagan (1978). A Bayesian prediction for the future world record is alsoconsidered
关键词：Importance sampling; local regression; monotone decreasing func-;tion; monotone spline; Monte Carlo numerical integration
