QUANTIFICATION OF OCKHAM'S RAZOR BY VIOLATION OF THE LIKELIHOOD PRINCIPLE.
Abstract
For inference situations with an ancillary statistic, a quantifier, based on a significance level whose calculation involves the distribution of the ancillary statistic, is described which provides a quantification of Ockham's Razor. Its application is proposed for those cases where the method of maximum likelihood fails, due to 'overparametrization', or rather, inadequate experimentation. Examples described include (i) a curve fitting problem (ii) a problem involving the observable convolution of two individually unobservable binomial variables. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1966
- Accession Number
- AD0634768
Entities
People
- Mervyn Stone
Organizations
- University of Wisconsin–Madison