Abstract:The arm of this paper is to establish the strong law of large numbers (SLLN) of $m$-dependent random variables under the framework of sub-linear expectations. We establish the SLLN for a sequence of independent, but not necessarily identically distributed random variables. The study further extends the SLLN to $m$-dependent and stationary sequence of random variables with the condition $C_{\mathbb V}(|X_1|)<\infty$ which is the sufficient and necessary condition of SLLN in the case of independent and identically distributed random variables.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to establish the strong law of large numbers (SLLN) for m - dependent random variables under the sub - linear expectation framework. Specifically, the goals of the paper are as follows:
1. **Establish the strong law of large numbers for sequences of independent but not identically distributed random variables**: Under the sub - linear expectation framework, establish the strong law of large numbers for sequences of independent but not identically distributed random variables.
2. **Extend to m - dependent and stationary sequences of random variables**: Further extend the strong law of large numbers to m - dependent and stationary sequences of random variables, and prove its validity under the condition \( \hat{C}_V(|X_1|) < \infty \). This condition is a necessary and sufficient condition for the strong law of large numbers in the case of independent and identically distributed random variables.
3. **Handle the non - additivity of sub - linear expectations and the discontinuity of capacity**: Due to the non - additivity of sub - linear expectations and the discontinuity of capacity, the strong law of large numbers in the classical probability framework cannot be directly applied to the m - dependent case. Therefore, the paper overcomes these difficulties by using the technique of splitting the sum, treating the sum of large and small pieces separately.
4. **Prove the necessary conditions for the strong law of large numbers for m - dependent and identically distributed random variables**: Even in the case of m - dependent and identically distributed, the existence of Choquet expectation is still a necessary condition for the strong law of large numbers to hold.
### Main Results
1. **Theorem 3.1**: For a sequence of independent but not identically distributed random variables \(\{X_n; n \geq 1\}\) in the sub - linear expectation space, if there exists a sequence \(\{a_n\}\) satisfying \(1 \leq a_n \uparrow \infty\) and \(\sum_{n = 1}^{\infty}\frac{\hat{E}[X_n^2]}{a_n^2}<\infty\), then:
\[
\hat{V}^* \left( \limsup_{n \to \infty} \frac{S_n - \hat{E}[S_n]}{a_n}>0 \text{ or } \liminf_{n \to \infty} \frac{S_n - \hat{E}[S_n]}{a_n}<0 \right) = 0.
\]
2. **Theorem 3.2**: Suppose \(\hat{E}\) satisfies condition (CC), then there exists a probability measure \(P \in \mathcal{P}\) such that:
\[
P \left( \limsup_{n \to \infty} \frac{S_n - \hat{E}[S_n]}{a_n} = 0 \text{ and } \liminf_{n \to \infty} \frac{S_n - \hat{E}[S_n]}{a_n} = 0 \right) = 1.
\]
3. **Theorem 3.3**: For an m - dependent and linearly stationary sequence of random variables \(\{X_n; n \geq 1\}\) in the sub - linear expectation space, if \( \hat{C}_V(|X_1|) < \infty \) is satisfied, then there exist real numbers \(\mu, \mu\) such that \(\hat{E}[X_1] \leq \mu \leq \mu \leq \hat{E}[X_1]\), and:
\[
\mu=\lim_{n \to \infty} \hat{E}[S_n]/n, \quad \mu=\lim_{n \to \infty} \hat{E}[S_n]/n,
\]
\[
\hat{V}^* \left( \limsup_{n \to \infty} \frac{S_n}{n}>\mu \text{ or } \liminf_{n \to \infty} \frac{S_n}{n}<\mu \right) = 0.