|
| << previous | next >> |
Journal description
All volumes of the corresponding journal
All issues of the corresponding volume
All articles of the corresponding issues
Details for article 11 of 15 found articles
| Title: |
Robustness of homogeneity of variance tests for randomized complete block data |
| Author: |
Schaalje, G. Bruce Despain, Darrin J. |
| Appeared in: |
Communications in statistics |
| Paging: | Volume 25 (1966) nr. 4 pages 961-977 |
| Year: |
1966 |
| Contents: |
Numerous statistics are available for testing homogeneity of variance using one-way data, and several studies have examined and compared robustness properties of these statistics. For randomized complete block data, only a few statistics are available for testing homogeneity of variance and no extensive comparative studies of robustness of these statistics have been carried out. This paper discusses eight such test statistics, and uses Monte Carlo simulation to investigate statistical properties of these statistics under the normal distribution and four non-normal distributions. They were also examined under a mild and a strong block effect. The Brindley-Bradley (1985) statistic is a valid and powerful statistic under normality, the Wilcox (1989) statistic emerged as a valid and powerful statistic under all kinds of non-normality if the block effect is mild, and a Levene-type statistic suggested by Yitnosumarto and O'Neill 1986 is recommended for symmetric non-normal distributions if the block effect is strong. Application of the tests to a data set involving infant thermometer types is discussed. |
| Publisher: |
Taylor & Francis |
| Source file: |
Elektronische Wetenschappelijke Tijdschriften |
Details for article 11 of 15 found articles
| << previous | next >> |