Predictive Distributions for Detecting Departures from an Uncertain Baseline
A predictive distribution is useful for monitoring departures from baseline performance. Several possible definitions of predictive distribution are presented and compared. The Bayesian definition using the Jeffreys non informative prior is recommended as a compromise among the competing definitions.
KeywordsNegative Binomial Distribution Predictive Distribution Future Data Past Data Homogeneous Poisson Process
Unable to display preview. Download preview PDF.
- 1.Eide SA, Atwood CL, Youngblood R, Rasmuson DM. Monitoring and reporting industry performance for the initiating events cornerstone of safety. INEEL/EXT-03-01179, 2003Google Scholar
- 2.Atwood CL. Predictive distributions for Poisson and binomial counts. STATWOOD-2002/1.1., 2002Google Scholar
- 4.Johnson NL, Kotz S. Continuous univariate distributions — 2. Wiley, New York, 1970, Sections 26.1 and 24.2.Google Scholar
- 5.Knüsel L. The prediction problem as the dual form of the two-sample problem with applications to the Poisson and the binomial distribution. Am. Statistician, 48: 214–219, 1994Google Scholar