1: \begin{abstract}

2: This paper studies the asymptotic power of tests of sphericity against

3: perturbations in a single unknown direction as both the dimensionality

4: of the data and the number of observations go to infinity. We establish

5: the convergence, under the null hypothesis and contiguous alternatives,

6: of the log ratio of the joint densities of the sample covariance

7: eigenvalues to a Gaussian process indexed by the norm of the

8: perturbation. When the perturbation norm is larger than the

9: \textit{phase transition threshold} studied in Baik, Ben~Arous and

10: P{\'e}ch{\'e} [\textit{Ann. Probab.} \textbf{33}

11: (\citeyear{BaiBenPec05}) 1643--1697] the limiting process is

12: degenerate, and discrimination between the null and the alternative is

13: asymptotically certain. When the norm is below the threshold, the

14: limiting process is nondegenerate, and the joint eigenvalue

15: densities under the null and alternative hypotheses are mutually

16: contiguous. Using the asymptotic theory of statistical experiments, we

17: obtain asymptotic power envelopes and derive the asymptotic power for

18: various sphericity tests in the contiguity region. In particular, we

19: show that the asymptotic power of the Tracy--Widom-type tests is

20: trivial (i.e., equals the asymptotic size), whereas that of the

21: eigenvalue-based likelihood ratio test is strictly larger than the

22: size, and close to the power envelope.\looseness=-1

23: \end{abstract}