1: \begin{abstract}

2:   We investigate traces of powers of random matrices whose

3:   distributions are invariant under rotations (with respect to the

4:   Hilbert--Schmidt inner product) within a real-linear subspace of the

5:   space of $n\times n$ matrices.  The matrices we consider may be real or

6:   complex, and Hermitian, antihermitian, or general.  We use Stein's

7:   method to prove multivariate central limit theorems, with

8:   convergence rates, for these traces of powers, which imply central

9:   limit theorems for polynomial linear eigenvalue statistics.  In

10:   contrast to the usual situation in random matrix theory, in our

11:   approach general, nonnormal matrices turn out to be easier to study

12:   than Hermitian matrices.

13: \end{abstract}

14: