An Investigation of the Power of the Wald-Wolfowitz, Two Sample, Runs Test
Abstract
In the absence of information concerning underlying distributions of populations being sampled, it is difficult to apply parametric statistical tests without possibly violating assumptions under which these tests have been derived. As a result, parametric statistical tests may provide invalid information and results in erroneous conclusions related to samples under observation. This undesirable effect leads statisticians toward the use of non- parametric tests which are unconcerned with the specific form of the underlying distributions. By computer sampling, the paper investigates the power of the Wald-Wolfowitz runs test as it pertains to normal, uniform and triangular distributions. The power is found to be satisfactory when it is possible to obtain large samples for comparison.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1972
- Accession Number
- AD0743078
Entities
People
- William C. Heschl
Organizations
- Naval Postgraduate School