Abstract
Using the Coulomb Fluid method, this paper derives central limit theorems (CLTs) for linear spectral statistics of three “spiked” Hermitian random matrix ensembles. These include Johnstone’s spiked model (i.e., central Wishart with spiked correlation), non-central Wishart with rank-one non-centrality, and a related class of non-central \(F\) matrices. For a generic linear statistic, we derive simple and explicit CLT expressions as the matrix dimensions grow large. For all three ensembles under consideration, we find that the primary effect of the spike is to introduce an \(O(1)\) correction term to the asymptotic mean of the linear spectral statistic, which we characterize with simple formulas. The utility of our proposed framework is demonstrated through application to three different linear statistics problems: the classical likelihood ratio test for a population covariance, the capacity analysis of multi-antenna wireless communication systems with a line-of-sight transmission path, and a classical multiple sample significance testing problem.
Similar content being viewed by others
Notes
The paper [35] considered a more general model than the spiked model. Therein, CLT results were presented for linear spectral statistics, with the key quantities involving solutions to implicit equations.
Note that \(x\text { dB}=10\log _{10}x\). These are the typical units used for expressing SNR in wireless communication systems.
Note that under \(H_0\), \(\bar{\mu }_{\text {R}} = 0\).
That is, when performing the test, the same decision threshold was chosen, as based on the asymptotic Gaussian distribution under \(H_0\).
References
Ledoit, O., Wolf, M.: Honey, I shrunk the sample covariance matrix. J. Portf. Manag. 30(4), 110–119 (2004)
Rubio, F., Mestre, X., Palomar, D.P.: Performance analysis and optimal selection of large minimum variance portfolios under estimation risk. IEEE J. Sel. Topics Signal Process. 6(4), 337–350 (2012)
Dahirel, V., Shekhar, K., Pereyra, F., Miura, T., Artyomov, M., Talsania, S., Allen, T.M., Altfeld, M., Carrington, M., Irvine, D.J., Walker, B.D., Chakraborty, A.K.: Coordinate linkage of HIV evolution reveals regions of immunological vulnerability. Proc. Natl. Acad. Sci. 108(28), 11530–11535 (2011)
Quadeer, A.A., Louie, R.H., Shekhar, K., Chakraborty, A.K., Hsing, I., McKay, M.R.: Statistical linkage analysis of substitutions in patient-derived sequences of genotype 1a Hepatitis C virus non-structural protein 3 exposes targets for immunogen design. J. Virol. 88(13), 7628–7644 (2014)
Hoydis, J., ten Brink, S., Debbah, M.: Massive mimo in the ul/dl of cellular networks: how many antennas do we need? IEEE J. Sel. Areas Commun. 2(2), 160–171 (2013)
Matthaiou, M., Zhong, C., McKay, M.R., Ratnarajah, T.: Sum rate analysis of zf receivers in distributed mimo systems. IEEE J. Sel. Areas Commun. 2(2), 180–191 (2013)
Dyson, F.J., Mehta, M.L.: Statistical theory of the energy levels of complex systems. J. Math. Phys. 4(5), 701–712 (1963)
Beenakker, C.: Universality in the random-matrix theory of quantum transport. Phys. Rev. Lett. 70(8), 1155–1158 (1993)
Beenakker, C.: Random-matrix theory of mesoscopic fluctuations in conductors and superconductors. Phys. Rev. B 47(23), 15763–15775 (1993)
Beenakker, C.: Universilaty of Br Tzin and Zee’s spectral correlator. Nucl. Phys. B 422, 515–520 (1995)
Chen, Y., Lawrence, N.: On the linear statistics of Hermitian random matrices. J. Phys. A 31(4), 1141–1152 (1998)
Lytova, A., Pastur, L.: Central limit theorem for linear eigenvalue statistics of random matrices with independent entries. Ann. Probab. 37(5), 1778–1840 (2009)
Costin, O., Lebowitz, J.: Gaussian fluctuation in random matrices. Phys. Rev. Lett. 75(1), 69–73 (1995)
Diaconis, P., Evans, S.N.: Linear functional of eigenvalues of random matrices. Trans. Am. Math. Soc. 353(7), 2615–2633 (2001)
Zheng, S.: Central limit theorems for linear spectral statistics of large dimensional \(F\)-matrices. Ann. Inst. Henri Poincaré Probab. Stat. 48(2), 444–476 (2012)
Johansson, K.: On fluctuations of eigenvalues of random Hermitian matrices. Duke Math. J. 91, 151–204 (1998)
Bai, Z.D., Jiang, D., Yao, J.-F., Zheng, S.: Corrections to LRT on large-dimensional covariance matrix by RMT. Ann. Stat. 37(6B), 3822–3840 (2009)
Chen, Y., McKay, M.R.: Coulumb fluid, Painlevé transcendents, and the information theory of MIMO systems. IEEE Trans. Inform. Theory 58(7), 4594–4634 (2012)
Bai, Z.D., Jiang, D., Yao, J.-F., Zheng, S.: Testing linear hypotheses in high-dimensional regressions. Statistics 47(6), 1207–1223 (2013)
Anderson, G.W., Zeitouni, O.: A CLT for a band matrix model. Probab. Theory Relat. Fields 134(2), 283–338 (2005)
Johnstone, I.M.: On the distribution of the largest eigenvalue in principal components analysis. Ann. Stat. 29(2), 295–327 (2001)
Baik, J., Ben, G.: Arous, S. Péché, Phase transition of the largest eigenvalue for non null complex sample covariance matrices. Ann. Probab. 33(5), 1643–1697 (2005)
Baik, J., Silverstein, J.: Eigenvalues of large sample covariance matrices of spiked population models. J. Multivar. Anal. 97(6), 1382–1408 (2006)
Paul, D.: Asymptotics of sample eigenstructure for a large dimensional spiked covariance model. Stat. Sinica. 17(4), 1617–1642 (2007)
Bai, Z.D., Yao, J.-F.: Central limit theorems for eigenvalues in a spiked population model. Ann. Inst. Henri Poincaré Probab. Stat. 44(3), 447–474 (2008)
Rao Nadakuditi, R., Silverstein, J.W.: Fundamental limit of sample generalized eigenvalue based detection of signals in noise using relatively few signal-bearing and noise-only samples. IEEE J. Sel. Top. Signal Proc. 4(3), 468–480 (2010)
Bai, Z.D., Yao, J.-F.: On sample eigenvalues in a generalized spiked population model. J. Multivar. Anal. 106, 167–177 (2012)
Rao Nadakuditi, R., Edelman, A.: Sample eigenvalue based detection of high-dimensional signals in white noise using relatively few samples. IEEE Trans. Signal Process. 56(7), 2625–2638 (2008)
Torun, M.U., Akansu, A.N., Avellaneda, M.: Portfolio risk in multiple frequencies. IEEE Signal Process. Mag. 5(28), 61–71 (2011)
Couillet, R., Debbah, M.: Random Matrix Methods for Wireless Communications, 1st edn. Cambridge University Press, New York (2011)
Bianchi, P., Debbah, M., Maida, M., Najim, J.: Performance of statistical tests for single-source detection using random matrix theory. IEEE Trans. Inform. Theory 57(4), 2400–2419 (2011)
Couillet, R., Hachem, W.: Fluctuations of spiked random matrix models and failure diagnosis in sensor networks. IEEE Trans. Inform. Theory 59(1), 509–525 (2013)
Simon, S.H., Moustakas, A.L., Marinelli, L.: Capacity and character expansions: moment-generating function and other exact results for MIMO correlated channels. IEEE Trans. Inform. Theory 53(12), 5336–5351 (2006)
Wang, Q., Silverstein, J.W., Yao, J.-F.: A note on the CLT of the LSS for sample covariance matrix from a spiked population model. J. Multivar. Anal. 130, 194–207 (2014)
Bai, Z.D., Silverstein, J.W.: CLT for linear spectral statistics of large-dimensional sample covariance matrices. Ann. Probab. 32(1A), 553–605 (2004)
Passemier, D., Yao, J.-F., On estimation of the noise variance in high-dimensional probabilistic principal component analysis, submitted to J. R. Statist. Soc. Ser. B (2013) arXiv:1308.3890
Wang, Q., Yao, J.-F.: On the sphericity test with large-dimensional observations. Electron. J. Stat. 7, 2164–2192 (2013)
Onatski, A., Moreira, M.J., Hallin, M.: Asymptotic power of sphericity tests for high-dimensional data. Ann. Stat. 41(3), 1204–1231 (2013)
van der Vaart, A.W.: Asymptotic Statistics, Cambridge Series in Statistical and Probabilistic Mathematics. Cambridge University Press, Cambridge (1998)
Dyson, F.J.: Statistical theory of energy levels of complex systems I-III. J. Math. Phys. 3(1), 140–175 (1962)
Chen, Y., Manning, S.M.: Asymptotic level spacing of the Laguerre ensemble: a Coulomb fluid approach. J. Phys. A 27(11), 3615–3620 (1994)
Chen, Y., Manning, S.M.: Distribution of linear statistics in random matrix models (metallic conductance fluctuations). J. Phys. 6(16), 3039–3044 (1994)
Chen, Y., Ismail, M.E.: Thermodynamic relations of the Hermitian matrix ensembles. J. Phys. A 30(19), 6633–6654 (1997)
Basor, E., Chen, Y.: Perturbed Hankel determinants. J. Phys. A 38(47), 10101–10106 (2005)
Simon, S.H., Moustakas, A.L.: Crossover from conserving to lossy transport in circular random-matrix ensembles. Phys. Rev. Lett. 96(13), 136805 (2006)
Vivo, P., Majumdar, S.N., Bohigas, O.: Large deviations of the maximum eigenvalue in Wishart random matrices. J. Phys. A 40(16), 4317–4337 (2007)
Vivo, P., Majumdar, S.N., Bohigas, O.: Distributions of conductance and shot noise and associated phase transitions. Phys. Rev. Lett. 101(21), 216809 (2008)
Dean, D.S., Majumdar, S.N.: Extreme value statistics of eigenvalues of Gaussian random matrices. Phys. Rev. E 77(4), 041108 (2008)
Katzav, E., Castillo, I.P.: Large deviations of the smallest eigenvalue of the Wishart-Laguerre ensemble. Phys. Rev. E 82(4), 040104 (2010)
Kazakopoulos, P., Mertikopoulos, P., Moustakas, A.L., Caire, G.: Living at the edge: a large deviations approach to the outage MIMO capacity. IEEE Trans. Inform. Theory 57(4), 1984–2007 (2011)
Li, S., McKay, M.R., Chen, Y.: On the distribution of MIMO mutual information: An in-depth Painlevé based characterization. IEEE Trans. Inform. Theory 59(9), 5271–5296 (2013)
Chen, Y., Haq, N.S., McKay, M.: Random matrix models, double-time Painlevé equations, and wireless relaying. J. Math. Phys. 54(6), 063506 (2013)
Wang, D.: The largest eigenvalue of real symmetric, Hermitian and Hermitian self-dual random matrix models with rank one external source, part I. J. Stat. Phys. 146(4), 719–761 (2012)
Mo, M.Y.: Rank 1 real Wishart spiked model. Commun. Pure Appl. Math. 65(11), 1528–1638 (2012)
Dharmawansa, P.: Three problems related to the eigenvalues of complex non-central Wishart matrices with rank-1 mean, submitted to SIAM J. Matrix Anal. Appl. (2013) arXiv:1306.6566
James, A.T.: Distributions of matrix variates and latent roots derived from normal samples. Ann. Math. Stat. 35(2), 475–501 (1964)
Mehta, M.L.: Random Matrices. Pure and Applied Mathematics, 3rd edn. Elsevier/Academic Press, Amsterdam (2004)
Mauchly, J.W.: Significance test for sphericity of a normal \(n\)-variate distribution. Ann. Math. Stat. 11(2), 204–209 (1940)
Anderson, T.W.: An Introduction to Multivariate Statistical Analysis. Wiley Series in Probability and Statistics, 3rd edn. Wiley, Hoboken (2003)
Onatski, A., Moreira, M.J., Hallin, M.: Signal detection in high dimension: the multispiked case. Ann. Stat. 42(1), 225–254 (2014)
Gradshteyn, I., Ryzhik, I.: Table of Integrals, Series, and Products, Alan Jeffrey. Academic Press, Boston (2000)
Foschini, G.J., Gans, M.J.: On limits of wireless communications in a fading environment when using multiple antennas. Wirel. Pers. Commun. 6(3), 311–335 (1998)
Telatar, I.E.: Capacity of multi-antenna Gaussian channels. Eur. Trans. Commun. 10(6), 585–595 (1999)
Chiani, M., Win, M.Z., Zanella, A.: On the capacity of spatially correlated MIMO Rayleigh-fading channels. IEEE Trans. Inform. Theory 49(10), 2363–2371 (2003)
Smith, P.J., Roy, S., Shafi, M.: Capacity of MIMO systems with semicorrelated flat fading. IEEE Trans. Inform. Theory 49(10), 2781–2788 (2003)
Kammoun, A., Kharouf, M., Hachem, W., Najim, J., El Kharroubi, A.: On the fluctuations of the mutual information for non centered MIMO channels: the non Gaussian case. In: Signal Processing Advances in Wireless Communications (SPAWC) (2010), pp. 1–5
Hachem, W., Loubaton, P., Najim, J.: Deterministic equivalents for certain functionals of large random matrices. Ann. Appl. Probab. 17(3), 875–930 (2007)
Moustakas, A.L., Simon, S.H.: Random matrix theory of multi-antenna communications: the Rician channel. J. Phys. A 38(49), 10859–10872 (2005)
Taricco, G.: Asymptotic mutual information statistics of separately correlated Rician fading MIMO channels. IEEE Trans. Inform. Theory 54(8), 3490–3504 (2008)
Kang, M., Alouini, M.-S.: Capacity of MIMO Rician channels. IEEE Trans. Wirel. Commun. 5(1), 112–122 (2006)
Alfano, G., Lozano, A., Tulino, A.M., Verdú, S.: Mutual information and eigenvalue distribution of MIMO Ricean channels. In: IEEE International Symposium on Information Theory and Application (ISITA), pp. 1040–1045 (2004)
Jayaweera, S.K., Poor, H.V.: On the capacity of multiple-antenna systems in Rician fading. IEEE Trans. Wirel. Commun. 4(3), 1102–1111 (2005)
J. Hansen, H. Bolcskei, A geometrical investigation of the rank-\(1\) Rician MIMO channel at high SNR. In: IEEE International Symposium on Information Theory (ISIT), p. 64 (2004)
McKay, M.R., Collings, I.B.: General capacity bound for spatially correlated Rician MIMO channels. IEEE Trans. Inform. Theory 51(9), 3121–3145 (2005)
Cui, X.W., Zhang, Q.T., Feng, Z.M.: Generic procedure for tightly bounding the capacity of MIMO correlated Rician fading channels. IEEE Trans. Commun. 53(4), 890–898 (2005)
McKay, M.R., Collings, I.B.: Improved lower general bound for spatially-correlated Rician MIMO capacity. IEEE Commun. Lett. 10(3), 162–164 (2006)
McKay, M.R., Smith, P.J., Collings, I.B.: New properties of complex noncentral quadratic forms and bounds on MIMO mutual information. In: IEEE International Symposium on Information Theory (ISIT), pp. 1209–1213 (2006)
Jin, S., Gao, X., You, X.: On the ergodic capacity of rank-1 Ricean-fading MIMO channels. IEEE Trans. Inform. Theory 53(2), 502–517 (2007)
Moustakas, A.L., Simon, S.H., Sengupta, A.M.: Mimo capacity through correlated channels in the presence of correlated interferers and noise: a (not so) large n analysis. IEEE Trans. Inform. Theory 49(10), 2545–2561 (2003)
Tulino, A., Verdu, S.: Asymptotic outage capacity of multiantenna channels. In: IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Vol. 5, pp. 825–828 (2005)
Hachem, W., Khorunzhiy, O., Loubaton, P., Najim, J., Pastur, L.: A new approach for mutual information analysis of large dimensional multi-antenna channels. IEEE Trans. Inform. Theory 54(9), 3987–4004 (2008)
Lozano, A., Tulino, A.M., Verdú, S.: High-SNR power offset in multiantenna communication. IEEE Trans. Inform. Theory 51(12), 4134–4151 (2005)
Schatzoff, M.: Comparative robustness of six tests in multivariate analysis of variance. J. Am. Stat. Assoc. 61(314), 415–435 (1966)
Olson, C.L.: Comparative robustness of six tests in multivariate analysis of variance. J. Am. Stat. Assoc. 69(348), 894–908 (1974)
Walpole, R.E., Myers, R.H., Myers, S.L., Ye, K.: Probability and Statistics for Engineers and Scientists, 8th edn. Prentice Hall, New Jersey (2003)
Takemura, A.: Zonal Polynomials, Institute of Mathematical Statistics Lecture Notes-Monograph Series, 4. Institute of Mathematical Statistics, Hayward (1984)
Dyson, F.J.: An Ising ferromagnet with discontinuous long-range order. Commun. Math. Phys. 21(4), 269–283 (1971)
Marčenko, V.A., Pastur, L.A.: Distribution of eigenvalues in certain sets of random matrices. Mat. Sb. (N.S.) 72(114), 507–536 (1967)
Olver, F.W.: Asymptotics and Special Functions, 4th edn. A K Peters, Natick (1997)
Bleistein, N., Handelsman, R.A.: Asymptotic Expansions of Integrals, 2nd edn. Dover Publications Inc., New York (1986)
Bai, Z.D.: Convergence rate of expected spectral distributions of large random matrices. II. Sample covariance matrices. Ann. Probab. 21(2), 649–672 (1993)
Abramowitz, M., Stegun, I.A.: Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, National Bureau of Standards Applied Mathematics Series. U.S. Government Printing Office, Washington DC (1964)
Tweddle, I.: James Stirling’s Methodus differentialis, Sources and Studies in the History of Mathematics and Physical Sciences, Springer-Verlag. London Ltd., London (2003)
Slater, L.J.: Confluent Hypergeometric Functions. Cambridge University Press, New York (1960)
Acknowledgments
Thanks to Prof. Iain Johnstone at Stanford University for useful discussions in relation to the non-central multivariate \(F\) matrices (Model C) and for pointing out associated applications. The work of D. Passemier and M. R. McKay was supported by the Hong Kong Research Grants Council (RGC) under Grant number 616911. The work of Y. Chen was supported by the Macau Science Foundation Grant, under Grant number FDCT 077/2012/A3.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Statement and Proof of Lemma 2
Lemma 2
Let \(u>0\) and \(v>0\) such that \(nu+1>0\) and \(n(v-u)\notin \mathbb {N}\). Assume that \(\mathfrak {R}z>1\) and \(\gamma \ge 0\). As \(n\rightarrow \infty \), we have
where \(t(z)= \frac{\gamma z -v+\sqrt{(v-\gamma z)^2+4\gamma z u}}{2\gamma z}\).
Proof
Under the prescribed conditions on \(u\) and \(v\), we may use the following integral representation [94]
where the contour \(C\) starts at 0, traverses anti-clockwise around 1 and returns to 0. For large \(n\), the Laplace approximation yields
for which
and where \(t_0\) is the saddlepoint, which is the solution to
Thus, \(t_0\) must satisfy
with \(t_0 \notin \{0,1\)}. There is one solution which lies outside the contour for \(\mathfrak {R}z > 1\):
Furthermore, we have
so that, taking the root with the correct phase factor (see [89, Chap. 4] or [90, Chap. 7] for more details), we get
Substituting this quantity with (72) into (71) together with (70), we find the desired result (69). \(\square \)
Appendix 2: Equivalence Between \(\int _a^b f(x) \rho _2 (x, z)\mathrm {d}x\) and \(\int _a^b \text {Log}(z-x) \rho _1(x) \mathrm {d}x\) in (39)
We want to show that, for \(z \notin [a,b]\),
where
and
This identity is not straightforward and appears difficult to show directly using the above expressions. Thus, here we adopt an approach based on first showing that the derivative with respect to \(z\) of (73) and (74) are equal. First considering (73), we have
Now taking the derivative of (74), we get
which, after applying the identities (86) and (89), yields
Application of integration by parts to the last integral gives
which is the same as (75). So we have proved that
Now, note that with \(z\) such that \(\mathfrak {I}z=0\), as \(\mathfrak {R}z \rightarrow \infty \),
Plugging this expression into (73) gives
which tends to zero as \(\mathfrak {R}z \rightarrow \infty \). Furthermore, with
the expression in (74) becomes, upon interchanging the integrals,
Here, the first principal value integral is zero by (86), whilst the remaining terms tends to zero when \(z\) is such that \(\mathfrak {R}z \rightarrow \infty \) and \(\mathfrak {I}z=0\).
Consequently, taking \(z\) such that \(\mathfrak {R}z \rightarrow \infty \) and \(\mathfrak {I}z=0\) in (76), we find that the constant term is zero, thus proving the result.
Appendix 3: Useful Formulas
For the derivations of our results, we will require numerous integrals; these are summarized in (77–93). Note that for all definite integrals involving the variable \(t\), these are valid for \(\mathfrak {R}t >b\), while in all cases we assume \(0<a<b\).
Moreover, for \(z \in \mathbb {C}\) and \(b<1\),
and for \(z \in \mathbb {C}\),
with \(A=\sqrt{(z-a)(z-b)}\).
Equations (77–88) are given in [18], whilst (89) is a slight modification of (78), and (90) follows using (78) and (79). The expression (91) follows upon multiplying the numerator and the denominator of the integrand by \(\sqrt{(b-x)(x-a)}\), applying a partial fraction decomposition, then invoking (79), (80) and (86). For (92) and (93), the derivations are more involved, and we describe each in turn.
For (92), we use the parametrization
and the partial fraction decomposition
to arrive at
The last equation was obtained by invoking (78) and (89). From a further change of variable \(x=1- \lambda z\), we have
Now consider (93). In this case, we use the relation
to give
The first integral is given by (89), whereas the double integral is
The first term on the right-hand side above is obtained using (89), whereas the second is
using the first identity in [61, Eq. 2.266]. Combining the previous calculations, we get the result.
Rights and permissions
About this article
Cite this article
Passemier, D., McKay, M.R. & Chen, Y. Asymptotic Linear Spectral Statistics for Spiked Hermitian Random Matrices. J Stat Phys 160, 120–150 (2015). https://doi.org/10.1007/s10955-015-1233-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10955-015-1233-x