The Central Limit Theorem Mathematical Proof
The central limit theorem says that the sample means of a any probability distribution with the sample size large enough, the sample means follow a normal distribution. This is useful in a way that, in real life, we don't certainly know what distribution things follow, and we may use the central limit theorem to model real life random events as long as we take big enough samples.
Moment Generating Function(MGF)
Moment generating function is a way to describe a probability distribution
just like a probability density function(PDF) or a cumulative distribution
function(CDF) does.
It is intuitive that two probability distributions are the same if their PDFs
are the same.
What if their MGFs are the same?
Definition
An MGF is defined as Mx(t) = E[etx], and if we write
out its Taylor series expansion, we get
Therefore,
If we plug 0 into t, we get Mx(0) = E[e0] = 1.
If we take the first derivative of Mx(t) with respect to t,
we get
Then if we plug 0 into t, we get E[x].
So, we can get the nth moment of a probability distribution by taking
its nth derivative with respect to t and plug 0
into t.
Uniqueness
If two probability distributions have the same MGFs, they have same moments
of all order.
For example. two exact probability distributions have
the same mean and the same variance, mean being E[x] and the variance being
E[x2] - E[x]2, and their first and second moments
must be the same.
On the other hand, if two probability distributions are not the same,
there must exist a moment of order n being different.
Central limit theorem proof
To proof the central limit theorem, we need one property of MGF.
If x and y are independent:
Mx+y(t) = E[et(x+y)] =
E[etxety] = E[etx] E[ety] =
Mx(t) My(t)
Proof
Suppose we take samples of size n form a probability
distribution and treat the sample means as random variable, we can
prove that the sample means have a MGF of a normal distribution.
Let a random sample be
x1 + x2 + x3 +...+ xn
and its mean be S
S = (x1 + x2 + x3 +...+ xn) /
n
The mathematical derivation is as the following:




Comments
Post a Comment