Null distributions for largest eigenvalues in multivariate analysis
by Iain Johnstone
Abstract: The eigenvalues of Wishart matrices play a central role in classical multivariate analysis. A new impetus to approximate distribution results has come from methods that imagine the number of variables as large. We focus on the largest eigenvalue in particular, and first briefly review null distribution approximations in terms of the Tracy-Widom laws. The second part will focus on work in progress on concentration inequalities for the largest eigenvalue in the "two Wishart" case, such as canonical correlations.
For More Information: Guoqi Qian, email@example.com