Accession Number : AD0717932

Title :   Exact Intervals and Tests for Median When One 'Sample' Value Possibly an Outlier.

Descriptive Note : Technical rept.,

Corporate Author : SOUTHERN METHODIST UNIV DALLAS TEX DEPT OF STATISTICS

Personal Author(s) : Walsh,John E.

Report Date : 18 DEC 1970

Pagination or Media Count : 8

Abstract : Available are n observations (continuous data) that are believed to be a random sample. Desired are confidence intervals and significance tests for the population median. However, there is the possibility that either the largest or the smallest observation is an outlier. That is, the population yielding this observation differs from the population yielding the other n - 1 observations. If this happens, intervals and tests are desired for the median of the population yielding the n - 1 observations. Some analysis difficulties would be avoided if intervals and tests could be developed that simultaneously are applicable for all three of these situations. More specifically, a confidence coefficient, or significance level, has the same value for all three situations. It is found that two-sided intervals and tests based on two symmetrically located order statistics (not the largest and smallest) have this property. Also, some extensions are considered wherein each observation can be from a different population. (Author)

Descriptors :   (*STATISTICAL DISTRIBUTIONS, CONFIDENCE LIMITS), STATISTICAL TESTS, SAMPLING

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE