Skip to main content

Large Sample Distribution Theory

Convergence in Probability​

Definition​

The random variable XnX_n converges in probability to a constant cc if

lim⁔nā†’āˆžĀ ProbĀ (∣Xnāˆ’c∣>ε)=0\lim_{n\to \infty}\text{ Prob }(|X_n-c|>\varepsilon)=0

for any positive ε\varepsilon.

If XnX_n converges in probability to cc, then we write

plimĀ Xn=c.\text{plim }X_n=c.

Another way to write is the following

lim⁔nā†’āˆžXn→pc,\lim_{n\to\infty} {X}_n\xrightarrow{p}c,

or

Xn→pc.{X}_n\xrightarrow{p}c.

Example​

Let Xn∼ Exponential(n)X_n \sim \text{ Exponential}(n), show that Xn→p0{X}_n\xrightarrow{p}0.

Hint: CDF of Xn∼ Exponential(λ)X_n \sim \text{ Exponential}(\lambda) id given as

F(x;Ī»)={1āˆ’eāˆ’Ī»xx≄0,0x<0F(x;\lambda)= \begin{cases} 1- e^{-\lambda x} & x\geq 0, \\ 0 &x< 0 \\ \end{cases}

Solution:

Using the definition of convergence in probability, we have

lim⁔nā†’āˆžProb(∣Xnāˆ’0∣>ε)=lim⁔nā†’āˆžProb(Xn>ε)(sinceĀ Xn≄0)=lim⁔nā†’āˆžeāˆ’nε(sinceĀ Xn∼ Exponential(n))=0,āˆ€Īµ>0.\begin{align*} \lim_{n\to \infty}\text{Prob}(|X_n-0|>\varepsilon) &= \lim_{n\to \infty}\text{Prob}(X_n>\varepsilon) &\hspace{50px}\text{(since } X_n\geq0)\\ &= \lim_{n\to \infty}e^{-n\varepsilon} &\hspace{50px}\text{(since } X_n \sim \text{ Exponential}(n))\\ &=0, \forall \varepsilon>0. \end{align*}

Convergence in Distribution​

Definition​

A sequence of random variables X1,X2,X3,⋯X_1, X_2, X_3,\cdots converges in distribution to a random variable XX, shown by Xn→dXX_n\xrightarrow{d}X, if,

lim⁔nā†’āˆžFXn(x)=FX(x),\lim_{n\to \infty}F_{X_n}(x)=F_X(x),

for all xx at which FX(x)F_X(x) is continuous.

Limiting Distribution​

If XnX_n converges in distribution to XX, where FXn(x)F_{X_n}(x) is the CDF of XnX_n, then FX(x)F_X(x) is the limiting distribution of XnX_n.

Example​

Let X1,X2,X3,⋯X_1,X_2, X_3,\cdots be a sequence of random variable such that

FXn(x)={1āˆ’(1āˆ’1n)nxx≄00otherwiseF_{X_n}(x)= \begin{cases} 1-\Big(1-\frac{1}{n}\Big)^{nx} & x\geq 0 \\ 0 & \text{otherwise} \\ \end{cases}

Show that XnX_n converges in distribution to Exponential(1)\text{Exponential(1)}.

Solution:

Let X∼Exponential(1)X∼\text{Exponential(1)}.
For x≤0x\leq0, we have

FXn(x)=FX(x)=0,Ā forĀ n=2,3,4,⋯ .F_{X_n}(x)=F_X(x)=0, \hspace{15px} \text{ for n}=2,3,4,\cdots.

For x≄0x\geq0, we have

lim⁔nā†’āˆžFXn(x)=lim⁔nā†’āˆž(1āˆ’(1āˆ’1n)nx)=1āˆ’lim⁔nā†’āˆž(1āˆ’1n)nx=1āˆ’eāˆ’x=FX(x),āˆ€x.\begin{align*} \lim_{n\to \infty}F_{X_n}(x)&=\lim_{n\to \infty}\Big(1āˆ’\Big(1āˆ’\frac{1}{n}\Big)^{nx}\Big)\\ &=1āˆ’\lim_{n\to\infty}\Big(1āˆ’\frac{1}{n}\Big)^{nx}\\ &=1āˆ’e^{āˆ’x}\\ &=F_X(x), \forall x. \end{align*}

Thus, we conclude that Xn→dXX_n \xrightarrow{d}X.