Proof of the Weak Law of Large Numbers and its Generalization
Posted onViews:
Proof of the Weak Law of Large Numbers (WLLN) and its generalization, which relaxes the finite expectation condition to a milder tail condition. The proof employs Chebyshev's inequality to establish convergence in probability.
Weak Law of Large Numbers
(WLLN)
Let be a
sequence of independent and identically distributed (i.i.d.) random
variables with finite expected value . Define the sample
mean as:
Then, converges
to in probability as approaches infinity (denoted as ).
Formally, for any , we have:
Generalized WLLN
Let be a
sequence of i.i.d. random variables such that
Then , where
Why it is a generalization?
Before proving the generalized version, let's see why it is indeed a
generalization of the standard WLLN. We first show that, for any random
variable , finite expected value
implies the condition required by the generalized WLLN:
Let ,
then
As , converges to almost surely. Combining this with the
fact that , which is
integrable, we can apply the Dominated Convergence Theorem to get:
Squeeze Theorem then gives us:
We also have as a direct consequence of DCT. Hence, if the
generalized WLLN holds, then the standard WLLN also holds.
To complete the discussion and assist understanding, we show that the
converse is not necessarily true (i.e., the condition of the generalized
WLLN is strictly weaker than the finite expectation condition). For
example, consider a random variable satisfying:
Then
but
Proof of the Generalized
WLLN
Proof Sketch
Denote
Our proof will proceed in two steps: - Show that . - Show that , where .
Step 1:
We consider when . This occurs if and only if there exists some
such that . Thus,
Since , the Squeeze Theorem gives us:
which directly implies .
Step 2:
For any , we must
show that
where
We will use Chebyshev's inequality to achieve this. Let , then