Level: BSc, MSci, MSc
Title: The sequential probability ratio test
Supervisor:

Dr Steve Coad
Research Area: Probability and Applications [Including Statistics]
Description:

The sequential probability ratio test (SPRT) is designed to decide between two simple hypotheses. The test decides in favour of one of these hypotheses on the basis of observations of the random variable under consideration, subject to certain error probabilities. When the number of observations is fixed, the Neyman–Pearson lemma shows that the most powerful test depends on the likelihood ratio. When observations become available sequentially, it was shown by Wald in the 1940s that the optimum test is the SPRT, which is a sequential analogue of the fixed-sample test. The project involves the study of some properties of the SPRT, such as the operating characteristic function, the average sample number and sample size distribution. Although the SPRT was developed in the context of industrial control, it has also been applied in clinical trials.
Further Reading:
  • S. Silvey, Statistical inference. Chapman & Hall, London, 1975.
  • A. Wald, Sequential analysis. Wiley, Chichester, 1947.
Key Modules:

MTH5121 Probability Models

MTH6136 Statistical Theory
Other Information:
Current Availability: Yes