A0597
Title: Correlation tests for high-dimensional noisy data
Authors: Aki Ishii - Tokyo University of Science (Japan) [presenting]
Yumu Iwana - University of Tsukuba (Japan)
Kazuyoshi Yata - University of Tsukuba (Japan)
Makoto Aoshima - University of Tsukuba (Japan)
Abstract: A correlation test is considered in high-dimension, low-sample-size scenarios, where the sample size is much smaller than the dimension under the strongly spiked eigenvalue (SSE) model. The test statistic given by a prior study is considered, and it is shown that an asymptotic distribution of the test statistic under the SSE model can be written by the distribution of the sum of weighted chi-squared variables. A new test procedure is proposed by using the asymptotic distribution. It is also shown that the proposed test procedure has preferable properties for size and power. Simulation studies and a demonstration with actual data analyses are also given using microarray data sets.
