Scholarlink

<p>When testing means of multivariate matrix-valued observations, good estimators of the variance matrix play a key role. If the design of experiment or nature of the data suggest that there can be present a special structure of the variance matrix, it is important to take it into account, since reduced number of parameters can significantly increase the power of the test. Recently, basic location tests have been derived for model with block compound symmetric (BCS) covariance structure under normality, improving upon standard Hotelling tests. Here, we add several variants of MANOVA tests for such situation. Moreover, we show that the tests work equally under all elliptically contoured distributions. The powers of the tests are compared via simulations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multiple testing of mean values in multivariate data with the BCS covariance structure

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月6日

查看全文
  • Ivan Žežula,
  • Daniel Klein

摘要

When testing means of multivariate matrix-valued observations, good estimators of the variance matrix play a key role. If the design of experiment or nature of the data suggest that there can be present a special structure of the variance matrix, it is important to take it into account, since reduced number of parameters can significantly increase the power of the test. Recently, basic location tests have been derived for model with block compound symmetric (BCS) covariance structure under normality, improving upon standard Hotelling tests. Here, we add several variants of MANOVA tests for such situation. Moreover, we show that the tests work equally under all elliptically contoured distributions. The powers of the tests are compared via simulations.