Let $a_n\geq 0$ be a sequence of non-negative numbers. Consider the following two statements: $$ \text{(I)}\qquad\qquad \lim_{n\to\infty} \frac{1}{n^2}\sum_{i=1}^n a_i =0, $$ $$ \text{(II)}\qquad\qquad\qquad \sum_{n=1}^\infty \frac{a_n}{n^2}<\infty. $$

Questions: Does (I) imply (II)? Does (II) imply (I)? Otherwise, please provide counterexamples.

Motivation: Both statements occur in the context of the law of large numbers for non-identically distributed random variables. With $a_n=\mathrm{Var}(X_n)$, one can conclude the weak LLN if the $X_n$ are pairwise uncorrelated and condition (I) holds. The strong LLN can be concluded if the $X_n$ are stochastically independent and condition (II) holds. Therefore, one might expect that (II) implies (I).


II implies I:

(we deal with the partial sums here)

One applies the Cauchy-Schwarz inequality to get:

$\displaystyle \left( \sum_{i=1}^n \frac{a_i}{i^2} \right)\left(\sum_{i=1}^n i^2\right) \ge (\sum_{i=1}^n a_i)^2 $

We can see that $ \sum_{i=1}^n i^2 $ is of the magnitude $n^3$ as $n \rightarrow \infty$. So when we divide both sides by $n^4$, LHS will converge to 0 as $n \rightarrow 0$, which implies that $\displaystyle \frac{\sum a_n}{n^2}$ converges to 0.

I does NOT imply II:

e.g. take $a_1,a_2$ to be anything, $\displaystyle a_n = \frac{n}{\log n}$ for $n \ge 3$. This does not satisfy II (well known). But it does satisfy I, because asymptotically, $\displaystyle \frac{1}{n^2} \sum_{i=1}^n a_i \sim \frac{1}{n^2} \int_3^n \frac{x}{\log x} dx$.

The logarithmic integral has a well known approximation $\displaystyle \int_3^n \frac{1}{\log x}dx = O\left(\frac{n}{\log n}\right)$.

So $\displaystyle \frac{1}{n^2} \int_3^n \frac{x}{\log x} dx \leq \frac{n}{n^2} \int_3^n \frac{1}{\log x} dx = O\left(\frac{1}{\log n}\right) \rightarrow 0$ as $n \rightarrow \infty$.

  • $\begingroup$ Nice! Thank you for your help. $\endgroup$ – Rasmus Aug 8 '10 at 13:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.