Statistical Inference with Delayed Data.
Abstract
Problems of inference from data accumulating at the random times new species are discovered were studied. Procedures for estimating a variety of characteristics of the population and for predicting the random probability that a next search will uncover a new species were developed. A death process was interposed and estimates of the size of a population and the mean life of its members were obtained. Results for a one-armed bandit in the presence of concomitant information were developed. These results apply to the sequential allocaiton of treatments in medical trials. Optimal and adaptive stopping based on the maximum of a sequence of dependent observations were studied. The results are likely to have application to the choice of components in redundant systems and to adaptive quality control. Methods have been derived to estimate the true significance level resulting from repeated significance tests. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 30, 1979
- Accession Number
- ADA069305
Entities
People
- Michael B. Woodroofe
- Norman Starr
Organizations
- University of Michigan