Cluster sets for partial sums and partial sum processes

Abstract: Let

X, X_{1}, X_{2}, l d o t s

be i.i.d. mean zero random vectors with values in a separable Banach space

B

,

S_{n} = X_{1} + c d o t s + X_{n}

for

n g e 1

, and assume

c_{n} : n g e 1

is a suitably regular sequence of constants. Furthermore, let

S_{(n)} (t), 0 l e t l e 1

be the corresponding linearly interpolated partial sum processes. We study the cluster sets

A = C (S_{n} / c_{n})

and

m a t h c a l A = C (S_{(n)} (c d o t) / c_{n})

. In particular,

A

and

m a t h c a l A

are shown to be nonrandom, and we derive criteria when elements in

B

and continuous functions

f : [0, 1] o B

belong to

A

and

m a t h c a l A

, respectively. When

B = m a t h b b R^{d}

we refine our clustering criteria to show both

A

and

m a t h c a l A

are compact, symmetric, and star-like, and also obtain both upper and lower bound sets for

m a t h c a l A

. When the coordinates of

X

in

m a t h b b R^{d}

are independent random variables, we are able to represent

m a t h c a l A

in terms of

A

and the classical Strassen set

m a t h c a l K

, and, except for degenerate cases, show

m a t h c a l A

is strictly larger than the lower bound set whenever

d g e 2

. In addition, we show that for any compact, symmetric, star-like subset

A

of

m a t h b b R^{d}

, there exists an

X

such that the corresponding functional cluster set

m a t h c a l A

is always the lower bound subset. If

d = 2

, then additional refinements identify

m a t h c a l A

as a subset of

(x_{1} g_{1}, x_{2} g_{2}) : (x_{1}, x_{2}) i n A, g_{1}, g_{2} i n m a t h c a l K

, which is the functional cluster set obtained when the coordinates are assumed to be independent.

Review text

Let \((X_n)_{n\geq1}\) be i.i.d.\ random vectors on \(\mathbb R^d\) with mean zero, partial sums \(S_n=\sum_{k=1}^nX_k\) and linearly interpolated partial sum process \(\{S_{(n)}(t)\}_{t\in[0,1]}\). As a continuation of their earlier works, the authors study the structure of the cluster sets \(A\) of \((c_n^{-1}S_n)_{n\geq1}\) and \(\mathcal A\) of \((c_n^{-1}\{S_{(n)}(t)\}_{t\in[0,1]})_{n\geq1}\) for a suitable sequence \((c_n)_{n\geq1}\) of positive constants. Their main theorem states that, if \(c_n^{-1}\sqrt{n}\to 0\), \(c_n/c_m\leq(1+\varepsilon)n/m\) for large \(m<n\), \(\sum_{n=1}^\infty \operatorname{P}\{\|X_1\|>c_n\}<\infty\) and \(\limsup_{n\to\infty}c_n^{-1}\|S_n\|<\infty\), then with probability one \(\mathcal A\) is a compact subset of \(C[0,1]\) and \(A\) is a compact in \(\mathbb R^d\), symmetric with respect to the origin and star-like. Moreover, upper and lower bound sets for \(\mathcal A\) are provided. An example is given such that, for an arbitrary compact, symmetric and star-like subset \(A\) of \(\mathbb R^d\), the corresponding functional cluster set \(\mathcal A\) coincides with the lower bound set. In the case where the coordinates of \(X_1\) are independent, a full characterization of \(\mathcal A\) in terms of \(A\) and the cluster set in Strassen's functional LIL for real-valued random variables is given. Preliminary to their main results, the authors first consider general cluster sets for the functional LIL of random variables with values in a separable Banach space \(B\) and derive criteria for a function \(f:[0,1]\to B\) to belong to the cluster set \(\mathcal A\). These criteria can be written in terms of Brownian motion probabilities in the special case of \(B=\mathbb R^d\) using a strong approximation result of \textit{A. I. Sakhanenko} [Prog. Probab. 47, 223--245 (2000; Zbl 0971.60032)], which enables the authors to prove their main results.

Recommendations

Cites work

Cited in

(6)

This page was built for publication: Cluster sets for partial sums and partial sum processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2450249)