On the strong law of large numbers and additive functions

  • PDF / 144,853 Bytes
  • 12 Pages / 595 x 842 pts (A4) Page_size
  • 27 Downloads / 209 Views

DOWNLOAD

REPORT


ON THE STRONG LAW OF LARGE NUMBERS AND ADDITIVE FUNCTIONS ´n Berkes1 , Wolfgang Mu ¨ller2 and Michel Weber3 Istva 1

Graz University of Technology, Institute of Statistics M¨ unzgrabenstrasse 11, A-8010 Graz, Austria E-mail: [email protected]

2

Graz University of Technology, Institute of Statistics M¨ unzgrabenstrasse 11, A-8010 Graz, Austria E-mail: [email protected] 3

IRMA, Universit´e Louis-Pasteur et C.N.R.S. 7 rue Ren´e Descartes, 67084 Strasbourg Cedex, France E-mail: [email protected] (Received September 24, 2010; Accepted October 29, 2010)

Dedicated to Endre Cs´ aki and P´ al R´ev´esz on the occasion of their 75th birthdays

Abstract Let f (n) be a strongly additive complex-valued arithmetic function. Under mild conditions on f , we prove the following weighted strong law of large numbers: if X, X1 , X2 , . . . is any sequence of integrable i.i.d. random variables, then PN n=1 f (n)Xn lim = EX a.s. PN N →∞ n=1 f (n)

1. Introduction Let X, X1 , X2 , . . . be i.i.d. integrable random variables and f (n), n = 1, 2, . . . Pn a positive numerical sequence, F (n) = k=1 f (k). By a classical result of Jamison, Orey and Pruitt [5], under the condition 1 lim sup #{n : F (n) ≤ xf (n)} < ∞, x→∞ x

(1)

Mathematics subject classification numbers: 60F15,11A25. Key words and phrases: strong law of large numbers, weighted i.i.d. sums, strongly additive functions. 1 Research supported by FWF grant S9603-N23 and OTKA grants K 67961 and K 81928. 0031-5303/2011/$20.00

c Akad´emiai Kiad´o, Budapest

Akad´ emiai Kiad´ o, Budapest Springer, Dordrecht

2

¨ I. BERKES, W. MULLER and M. WEBER

we have the weighted strong law Pn k=1 f (k)Xk lim P = EX n n→∞ k=1 f (k)

a.s.

(2)

Conversely, if (2) holds for all i.i.d. sequences X, X1 , X2 , . . . with finite means, then (1) is valid. Note that condition (1) puts a restriction on the distribution of the weight sequence f (n) and not on the magnitude of the weights, as it happens, e.g., in central limit theory. In particular, (1) can fail even for bounded weight sequences f (n), see [5]. Condition (1) is generally difficult to check for irregular sequences f (n), and this leads to the question to study (2) for typical irregular sequences in number theory, for example, additive arithmetic functions. Let f (n), n = 1, 2, . . ., be a real-valued, strongly additive function, i.e., assume that f (mn) = f (m) + f (n)

for (m, n) = 1

(3)

and f (pα ) = f (p),

for p prime, α = 2, 3, . . .

It follows that f (n) =

X

(4)

f (p),

p|n

so that f is completely determined by its values taken over the primes. A typical example is ω(n), the number of different prime factors of n. Put An =

X f (p) , p

Bn =

p≤n

X |f (p)|2 . p

(5)

p≤n

In [1] we studied the weighted SLLN with coefficients f (n) and proved the following result (see Theorem 1.1 in [1]). Theorem 1. Assume that f ≥ 0 and Bp → ∞,

f (p) = o(Bp1/2 )

as p → ∞.

(6)

Then for any i.i.d. sequence X, X1 , X2 , . . . with finite means, the weighted strong law (2) holds. Condition (6) plays an important role in probabi