Kinds of Polyconfidence Intervals for Centers, and Some Thoughts on Identification and Selection of Confidence Procedures Using Configural Polysampling.
Abstract
If we deal initially only with single situations, monoconfidence intervals can be either required only to be weak, to have proper confidence when averaged over all configurations, or to be strong, when they must have the appropriate properties conditionally for each configuration. Assuming location-and-scale invariance throughout both for situations and confidence intervals. When one considers a plot involving several situations, three kinds of polyconfidence intervals seem worth mention: (A) doubly-strong, whose properties hold separately for each combination of a situation and a configuration, (b) weak, whose properties hold on average for each situation, (c) singly strong, where the properties hold, first on average for each situation and, second, for each configuation and one situation. These notions, as well as that of balance in a conservative sense, are explored in the framework of configurations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 16, 1981
- Accession Number
- ADA105907
Entities
People
- John W. Tukey
Organizations
- Princeton University