Statistics Seminars: Nonparametric predictive inference for multinomial data
13 May 2005 00:00 in CM107
"We present a new interval probabilistic method for predictive inference based on multinomial data. The underlying model uses a probability wheel representation with segments representing observation categories, and an adapted version of Hill's assumption A_(n), which is closely related to exchangeability, to link future observations to data. We compare this approach to Walley's `Imprecise Dirichlet Model' (IDM). (This is joint work with Thomas Augustin (Munich).) "
Contact email@example.com for more information