Comparison of exact parametric tests for high-dimensional data

Several proposals for exact or at least conservative parametric multivariate tests in the general linear model are considered that are applicable also for high-dimensional data, where the dimension of the observations may exceed the sample size. The common feature is the inclusion of principal component transformations into the test. Whereas a test proposal by Srivastava and von Rosen [Srivastava, M.S., von Rosen, D., 2004. MANOVA with singular variance matrix. Acta et Commentationes Universitatis Tartuensis de Mathematica 8, 253-269] originally assumes that the multivariate data have a known reduced rank which is used in the construction of the test, several versions of so-called PC tests by Lauter and colleagues accept a reduction of variance and utilize it for a ''stabilization'' of the test in terms of power. The different tests are compared with respect to their philosophy as well as in their performance in two real data examples and in simulation studies. It is shown that the test of Srivastava and von Rosen is a conservative test, even when a rank is assumed that is smaller than the true one. It is, however, less conservative than the conservative version of the PC test derived for the construction of a convenient confidence region for the investigated effects. The exact PC tests turns out to have the largest power.

[1]  A. Moore,et al.  Relationships between subtropical, semiarid forest of Acacia harpophylla (Brigalow), microrelief, and chemical properties of associated gilgai soil , 1967 .

[2]  Arjun K. Gupta,et al.  On the exact distribution of Wilks's criterion , 1969 .

[3]  S. Kropf.  Confidence regions for stabilized multivariate tests , 2000 .

[4]  T. Caliński,et al.  Linear Statistical Inference , 1985 .

[5]  Muni S. Srivastava,et al.  An introduction to applied multivariate statistics , 1984 .

[6]  J. Läuter,et al.  A theorem on the principal components inference , 2005 .

[7]  Øyvind Langsrud,et al.  Rotation tests , 2005, Stat. Comput..

[8]  Jurgen Lauter,et al.  Exact t and F Tests for Analyzing Studies with Multiple Endpoints , 1996 .

[9]  Search for relevant sets of variables in a high‐dimensional setup keeping the familywise error rate , 2005 .

[10]  Calyampudi Radhakrishna Rao,et al.  Linear Statistical Inference and its Applications , 1967 .

[11]  P. O'Brien.  Procedures for comparing samples with multiple endpoints. , 1984, Biometrics.

[12]  J M Lachin,et al.  Some large-sample distribution-free estimators and tests for multivariate partially incomplete data from two populations. , 1992, Statistics in medicine.

[13]  M. Srivastava,et al.  Regression models with unknown singular covariance matrix , 2002 .

[14]  Jürgen Läuter,et al.  New multivariate tests for data with an inherent structure , 1996 .

[15]  Siegfried Kropf,et al.  New Tests for Data with an Inherent Structure , 1996 .

[16]  Calyampudi R. Rao,et al.  Linear Statistical Inference and Its Applications. , 1975 .

[17]  H. Heuer,et al.  Comparison of Independent Samples of High‐Dimensional Data by Pairwise Distance Measures , 2007, Biometrical journal. Biometrische Zeitschrift.

[18]  S. Kropf,et al.  Multivariate tests based on left-spherically distributed linear scores , 1998 .