
It is shown that the tessellation of a compact, hyperbolic surface induced by
a typical long geodesic segment, when properly scaled, looks locally like a
Poisson line process. This implies that the global statistics of the
tessellation  for instance, the fraction of triangles  approach those of
the limiting Poisson line process.

Voters making a binary decision purchase votes from a centralized clearing
house, paying the square of the number of votes purchased. The net payoff to an
agent with utility $u$ who purchases $v$ votes is $\Psi (S_{n+1})uv^{2}$,
where $\Psi$ is a monotone function taking values between 1 and +1 and
$S_{n+1}$ is the sum of all votes purchased by the $n+1$ voters participating
in the election. The utilities of the voters are assumed to arise by random
sampling from a probability distribution $F_{U}$ with compact support; each
voter knows her own utility, but not those of the other voters, although she
does know the sampling distribution $F_{U}$. Nash equilibria for this game are
described. These results imply that the expected inefficiency of any Nash
equilibrium decays like $1/n$.

We obtain sharp asymptotic estimates for hitting probabilities of a critical
branching Brownian motion in one dimension with killing at 0 We also obtain
sharp asymptotic formulas for the tail probabilities of the number of particles
killed at 0. In the special case of doubleornothing branching, we give exact
formulas for both the hitting probabilities, in terms of elliptic functions,
and the distribution of the number of killed particles.

We consider a branching random walk initiated by a single particle at
location 0 in which particles alternately reproduce according to the law of a
GaltonWatson process and disperse according to the law of a driftless random
walk on the integers. When the offspring distribution has mean 1 the branching
process is critical, and therefore dies out with probability 1. We prove that
if the particle jump distribution has mean zero, positive finite variance
$\eta^{2}$, and finite $4+\varepsilon$ moment, and if the offspring
distribution has positive variance $\sigma^{2}$ and finite third moment then
the distribution of the rightmost position $M$ reached by a particle of the
branching random walk satisfies $P\{M \geq x\}\sim 6\eta^{2}/
(\sigma^{2}x^{2})$ as $x \rightarrow \infty$. We also prove a conditional limit
theorem for the distribution of the rightmost particle location at time $n$
given that the process survives for $n$ generations.

We consider measurevalued processes $X=(X_t)$ that solve the following
martingale problem: for a given initial measure $X_0$, and for all smooth,
compactly supported test functions $\varphi$, \begin{eqnarray*}X_t(\varphi
)=X_0(\varphi)+\frac{1}{2}\int _0^tX_s(\Delta \varphi )\,ds+\theta
\int_0^tX_s(\varphi )\,ds\\{}\int_0^tX_s(L_s\varphi )\,ds+M_t(\varphi
).\end{eqnarray*} Here $L_s(x)$ is the local time density process associated
with $X$, and $M_t(\varphi )$ is a martingale with quadratic variation
$[M(\varphi )]_t=\int_0^tX_s(\varphi ^2)\,ds$. Such processes arise as scaling
limits of SIR epidemic models. We show that there exist critical values
$\theta_c(d)\in(0,\infty)$ for dimensions $d=2,3$ such that if
$\theta>\theta_c(d)$, then the solution survives forever with positive
probability, but if $\theta<\theta_c(d)$, then the solution dies out in finite
time with probability 1. For $d=1$ we prove that the solution dies out almost
surely for all values of $\theta$. We also show that in dimensions $d=2,3$ the
process dies out locally almost surely for any value of $\theta$; that is, for
any compact set $K$, the process $X_t(K)=0$ eventually.

Any exchangeable Markov processes on $[k]^{\mathbb{N}}$ with cadlag sample
paths projects to a Markov process on the simplex whose sample paths are cadlag
and of locally bounded variation. Furthermore, any such process has a de
Finettitype description as a mixture of i.i.d. copies of timeinhomogeneous
Markov processes on $[k]$. In the Feller case, these timeinhomogeneous Markov
processes have a relatively simple structure; however, in the nonFeller case a
greater variety of behaviors is possible since the transition law of the
underlying Markov process on $[k]^{\zz{N}}$ can depend in a nontrivial way on
the exchangeable $\sigma$algebra of the process.

We consider a critical continuoustime branching process (a Yule process) in
which the individuals independently execute symmetric $\alpha$stable random
motions on the real line starting at their birth points. Because the branching
process is critical, it will eventually die out, and so there is a welldefined
maximal location $M$ ever visited by an individual particle of the process. We
prove that the distribution of $M$ satisfies the asymptotic relation $P\{M\geq
x \}\sim (2/\alpha)^{1/2}x^{\alpha /2}$ as $x \rightarrow \infty$.

Let $\Upsilon $ be a compact, negatively curved surface. From the (finite)
set of all closed geodesics on $\Upsilon$ of length $\leq L$, choose one, say
$\gamma_{L}$, at random and let $N (\gamma_{L})$ be the number of its
selfintersections. It is known that there is a positive constant $\kappa$
depending on the metric such that $N (\gamma_{L})/L^{2} \rightarrow \kappa$ in
probability as $L\rightarrow \infty$. The main results of this paper concern
the size of typical fluctuations of $N (\gamma_{L})$ about $\kappa L^{2}$. It
is proved that if the metric has constant curvature 1 then typical
fluctuations are of order $L$, in particular, $(N (\gamma_{L})\kappa L^{2})/L$
converges weakly to a nondegenerate probability distribution. In contrast, it
is also proved that if the metric has variable negative curvature then
fluctuations of $N (\gamma_{L})$ are of order $L^{3/2}$, in particular, $(N
(\gamma_{L})\kappa L^{2})/L^{3/2}$ converges weakly to a Gaussian
distribution. Similar results are proved for generic geodesics, that is,
geodesics whose initial tangent vectors are chosen randomly according to
normalized Liouville measure.

We study the convergence rate to stationarity for a class of exchangeable
partitionvalued Markov chains called cutandpaste chains. The law governing
the transitions of a cutandpaste chain are determined by products of i.i.d.
stochastic matrices, which describe the chain induced on the simplex by taking
asymptotic frequencies. Using this representation, we establish upper bounds
for the mixing times of ergodic cutandpaste chains, and under certain
conditions on the distribution of the governing random matrices we show that
the "cutoff phenomenon" holds.

It is proved that the Green's function of a symmetric finite range random
walk on a cocompact Fuchsian group decays exponentially in distance at the
radius of convergence R. It is also shown that Ancona's inequalities extend to
R, and therefore that the Martin boundary for Rpotentials coincides with the
natural geometric boundary S^1, and that the Martin kernel is uniformly
H\"older continuous. Finally, it is proved that this implies a local limit
theorem for the transition probabilities.

Oriented closed curves on an orientable surface with boundary are described
up to continuous deformation by reduced cyclic words in the generators of the
fundamental group and their inverses. By selfintersection number one means the
minimum number of transversal selfintersection points of representatives of
the class. We prove that if a class is chosen at random from among all classes
of $m$ letters, then for large $m$ the distribution of the selfintersection
number approaches the Gaussian distribution.

It is proved that the Green's function of the simple random walk on a surface
group of large genus decays exponentially at the spectral radius. It is also
shown that Ancona's inequalities extend to the spectral radius R, and therefore
that the Martin boundary for Rpotentials coincides with the natural geometric
boundary S^1. This implies that the Green's function obeys a power law with
exponent 1/2 at the spectral radius.

We study the fluctuations of selfintersection counts of random geodesic
segments of length $t$ on a compact, negatively curved surface in the limit of
large $t$. If the initial direction vector of the geodesic is chosen according
to the \emph{Liouville measure}, then it is not difficult to show that the
number $N (t)$ of selfintersections by time $t$ grows like $\kappa t^{2}$,
where $\kappa =\kappa_{M}$ is a positive constant depending on the surface $M$.
We show that (for a smooth modification of $N (t)$) the fluctuations are of
size $t$, and the limit distribution is a weak limit of Gaussian quadratic
forms. We also show that the fluctuations of \emph{localized} selfintersection
counts (that is, only selfintersections in a fixed subset of $M$ are counted)
are typically of size $t^{3/2}$, and the limit distribution is Gaussian.

The behavior at criticality of spatial SIR (susceptible/infected/recovered)
epidemic models in dimensions two and three is investigated. In these models,
finite populations of size N are situated at the vertices of the integer
lattice, and infectious contacts are limited to individuals at the same or at
neighboring sites. Susceptible individuals, once infected, remain contagious
for one unit of time and then recover, after which they are immune to further
infection. It is shown that the measurevalued processes associated with these
epidemics, suitably scaled, converge, in the largeN limit, either to a
standard DawsonWatanabe process (superBrownian motion) or to a
DawsonWatanabe process with locationdependent killing, depending on the size
of the the initially infected set. A key element of the argument is a proof of
Adler's 1993 conjecture that the local time processes associated with branching
random walks converge to the local time density process associated with the
limiting superBrownian motion.

In the simple meanfield SIS and SIR epidemic models, infection is
transmitted from infectious to susceptible members of a finite population by
independent $p$coin tosses. Spatial variants of these models are proposed, in
which finite populations of size $N$ are situated at the sites of a lattice and
infectious contacts are limited to individuals at neighboring sites. Scaling
laws for both the meanfield and spatial models are given when the infection
parameter $p$ is such that the epidemics are critical. It is shown that in all
cases there is a critical threshold for the numbers initially infected: below
the threshold, the epidemic evolves in essentially the same manner as its
branching envelope, but at the threshold evolves like a branching process with
a sizedependent drift.

In the simple meanfield SIS and SIR epidemic models, infection is
transmitted from infectious to susceptible members of a finite population by
independent pcoin tosses. Spatial variants of these models are proposed, in
which finite populations of size N are situated at the sites of a lattice and
infectious contacts are limited to individuals at neighboring sites. Scaling
laws for these models are given when the infection parameter p is such that the
epidemics are critical. It is shown that in all cases there is a critical
threshold for the numbers initially infected: below the threshold, the epidemic
evolves in essentially the same manner as its branching envelope, but at the
threshold evolves like a branching process with a sizedependent drift. The
corresponding scaling limits are superBrownian motions and DawsonWatanabe
processes with killing, respectively.

When do nonparametric Bayesian procedures ``overfit''? To shed light on this
question, we consider a binary regression problem in detail and establish
frequentist consistency for a certain class of Bayes procedures based on
hierarchical priors, called uniform mixture priors. These are defined as
follows: let $\nu$ be any probability distribution on the nonnegative integers.
To sample a function $f$ from the prior $\pi^{\nu}$, first sample $m$ from
$\nu$ and then sample $f$ uniformly from the set of step functions from $[0,1]$
into $[0,1]$ that have exactly $m$ jumps (i.e., sample all $m$ jump locations
and $m+1$ function values independently and uniformly). The main result states
that if a datastream is generated according to any fixed, measurable
binaryregression function $f_0\not\equiv1/2$, then frequentist consistency
obtains: that is, for any $\nu$ with infinite support, the posterior of
$\pi^{\nu}$ concentrates on any $L^1$ neighborhood of $f_0$. Solution of an
associated largedeviations problem is central to the consistency proof.

This paper is concerned with the problem of recovering a finite,
deterministic time series from observations that are corrupted by additive,
independent noise. A distinctive feature of this problem is that the available
data exhibit longrange dependence and, as a consequence, existing statistical
theory and methods are not readily applicable. This paper gives an analysis of
the denoising problem that extends recent work of Lalley, but begins from first
principles. Both positive and negative results are established. The positive
results show that denoising is possible under somewhat restrictive conditions
on the additive noise. The negative results show that, under more general
conditions on the noise, no procedure can recover the underlying deterministic
series.

The Northeast Model is a spin system on the twodimensional integer lattice
that evolves according to the following rule: Whenever a site's southerly and
westerly nearest neighbors have spin $1$, it may reset its own spin by tossing
a $p$coin; at all other times, its spin remains frozen. It is proved that the
northeast model has a phase transition at $p_{c}=1\beta_{c}$, where
$\beta_{c}$ is the critical parameter for oriented percolation. For $p<p_{c}$,
the trivial measure $\delta_{0}$ that puts mass one on the configuration with
all spins set at $0$ is the unique ergodic, translation invariant, stationary
measure. For $p\geq p_{c}$, the product Bernoulli$p$ measure on configuration
space is the unique nontrivial, ergodic, translation invariant, stationary
measure for the system, and it is mixing. For $p>2/3$ it is shown that there is
exponential decay of correlations.

We consider a twotype stochastic competition model on the integer lattice
Z^d. The model describes the space evolution of two ``species'' competing for
territory along their boundaries. Each site of the space may contain only one
representative (also referred to as a particle) of either type. The spread
mechanism for both species is the same: each particle produces offspring
independently of other particles and can place them only at the neighboring
sites that are either unoccupied, or occupied by particles of the opposite
type. In the second case, the old particle is killed by the newborn. The rate
of birth for each particle is equal to the number of neighboring sites
available for expansion. The main problem we address concerns the possibility
of the longterm coexistence of the two species. We have shown that if we start
the process with finitely many representatives of each type, then, under the
assumption that the limit set in the corresponding first passage percolation
model is uniformly curved, there is positive probability of coexistence.