English

Phase transition in random contingency tables with non-uniform margins

Probability 2020-09-15 v5 Combinatorics Statistics Theory Statistics Theory

Abstract

For parameters n,δ,B,n,\delta,B, and CC, let X=(Xk)X=(X_{k\ell}) be the random uniform contingency table whose first nδ\lfloor n^{\delta} \rfloor rows and columns have margin BCn\lfloor BCn \rfloor and the last nn rows and columns have margin Cn\lfloor Cn \rfloor. For every 0<δ<10<\delta<1, we establish a sharp phase transition of the limiting distribution of each entry of XX at the critical value Bc=1+1+1/CB_{c}=1+\sqrt{1+1/C}. In particular, for 1/2<δ<11/2<\delta<1, we show that the distribution of each entry converges to a geometric distribution in total variation distance, whose mean depends sensitively on whether B<BcB<B_{c} or B>BcB>B_{c}. Our main result shows that E[X11]\mathbb{E}[X_{11}] is uniformly bounded for B<BcB<B_{c}, but has sharp asymptotic C(BBc)n1δC(B-B_{c}) n^{1-\delta} for B>BcB>B_{c}. We also establish a strong law of large numbers for the row sums in top right and top left blocks.

Keywords

Cite

@article{arxiv.1903.08743,
  title  = {Phase transition in random contingency tables with non-uniform margins},
  author = {Sam Dittmer and Hanbaek Lyu and Igor Pak},
  journal= {arXiv preprint arXiv:1903.08743},
  year   = {2020}
}

Comments

24 pages, 4 figures. Earlier version is accepted for publication in Transactions of the AMS. This version contains an appendix on asymptotic independence of the entries