How many entries of a typical orthogonal matrix can be approximated by independent normals?

DOI10.1214/009117906000000205zbMATH Open1107.15018arXivmath/0601457OpenAlexW2089817321MaRDI QIDQ850976FDOQ850976

Publication date: 8 November 2006

Published in: The Annals of Probability (Search for Journal in Brave)

Abstract: We solve an open problem of Diaconis that asks what are the largest orders of

p_{n}

and

q_{n}

such that

Z_{n},

the

p_{n} i m e s q_{n}

upper left block of a random matrix

which is uniformly distributed on the orthogonal group O(n), can be approximated by independent standard normals? This problem is solved by two different approximation methods. First, we show that the variation distance between the joint distribution of entries of

Z_{n}

and that of

p_{n} q_{n}

independent standard normals goes to zero provided

p_{n} = o (s q r t n)

and

q_{n} = o (s q r t n)

. We also show that the above variation distance does not go to zero if

p_{n} = [x s q r t n]

and

q_{n} = [y s q r t n]

for any positive numbers

x

and

y

. This says that the largest orders of

p_{n}

and

q_{n}

are

o (n^{1 / 2})

in the sense of the above approximation. Second, suppose

is generated by performing the Gram--Schmidt algorithm on the columns of

, where

y_{i j}; 1 l e q i, j l e q n

are i.i.d. standard normals. We show that

e p s i l o n_{n} (m) := m a x_{1 l e q i l e q n, 1 l e q j l e q m} | s q r t n c d o t g a m m a_{i j} - y_{i j} |

goes to zero in probability as long as

m = m_{n} = o (n / l o g n)

. We also prove that

e p s i l o n_{n} (m_{n}) o 2 s q r t a l p h a

in probability when

m_{n} = [n a l p h a / l o g n]

for any

a l p h a > 0.

This says that

m_{n} = o (n / l o g n)

is the largest order such that the entries of the first

m_{n}

columns of

can be approximated simultaneously by independent standard normals.

Full work available at URL: https://arxiv.org/abs/math/0601457

Recommendations

zbMATH Keywords

random matrix large deviation Haar measure

Mathematics Subject Classification ID

Convergence of probability measures (60B10) Multivariate distribution of statistics (62H10) Random matrices (algebraic aspects) (15B52) Central limit and other weak theorems (60F05) Probability measures on groups or semigroups, Fourier transforms, factorization (60B15)

Cites Work

Cited In (37)

This page was built for publication: How many entries of a typical orthogonal matrix can be approximated by independent normals?

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q850976)