Abstract
A closed sequential binomial test has finite length, and the determination of its statistical properties takes this into account. This contrasts with Wald sequential tests for which the properties are based on infinite length. Wald tests can be truncated, but theory provides no information on the effect of truncation. Tests truncated at two or three times the average sample number are assumed to maintain their statistical properties. This assumption is inappropriate for short tests. A BASIC computer program is provided that will calculate exact statistical properties of closed (or truncated) sequential binomial tests. This program can be used iteratively to design sequential binomial sign tests of any length.
Additional information
Notes on contributors
Lloyd S. Nelson
Dr. Nelson is a private statistical consultant. He is a Fellow of ASQC.