SOME BOUNDED SIGNIFICANCE LEVEL TESTS FOR THE MEDIAN
Abstract
In practice it is often permissible to assume that the observations of a set are statistically independent and from continuous populations with a common median. Then the population median can be investigated by using the sign test. For small numbers of observations, however, the sign test does not furnish very many suitable significance levels. Also, some of the sign tests with suitable significance levels are not very efficient. The note presents some tests whose significance levels are only approximate but cover a wide range of suitable values. The significance levels of these tests are exactly determined if the populations are symmetrical; they are bounded otherwise. Some of these bounded significance level tests have high efficiencies.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 16, 1950
- Accession Number
- AD0603882
Entities
People
- John Walsh
Organizations
- RAND Corporation