Title: Asymptotic robustness for error rate of 2 group discriminant analysis for large dimensional case
Authors: Takayuki Yamada - Shimane university (Japan) [presenting]
Tetsuro Sakurai - Suwa university of science (Japan)
Yasunori Fujikoshi - Hiroshima university (Japan)
Abstract: The focus is on the problems for 2-groups linear discriminant analysis for high-dimensional data when the covariance matrices are equal. Firstly, we show that the asymptotic approximation of the error rate under non-normality as the dimension and sample size go to infinity together. An asymptotic estimator for the error rate is also obtained under the above asymptotic framework. A small-scaled simulation is carried out to confirm the precision of the approximation. We also show an asymptotic approximation for the error rate of the unified type discriminant statistic which includes linear discriminant function and quadratic discriminant function.