MIT15 075JF11 chpt05

Central Limit Theorem
(Convergence of the sample mean’s distribution to the normal distribution)

Let X1, X2, . . . , Xn be a random sample drawn from any distribution with a finite mean µ
and variance σ2. As n → ∞, the distribution of:
X̄ − µ
√
σ/ n
“converges” to the distribution N (0, 1). In other words,
X̄ − µ
√ ≈ N (0, 1).
σ/ n
X̄ −
√µ ?
Note 1: What is Remember that we proved that E(X̄ ) = µ and Var(X̄ ) = σ 2 /n.
σ/ n
That means we are taking the random variable X̄ , subtracting its mean, and dividing by its
standard deviation. It’s a z-score!
Note 2: “converge” means “convergence in distribution:”

X̄ − µ
lim P z = Φ(z) for all z.
n→∞
√ σ/ n ≤
Don’t worry about this if you don’t understand (it’s beyond the scope of 15.075).
Note 3: CLT is really useful because it characterizes large samples from any distribution.
As long as you have a lot of independent samples (from any distribution), then the distribu-
tion of the sample mean is approximately normal.
Let’s demonstrate the CLT.

Pick n large. Draw n observations from U [0, 1] (or whatever distribution you like). Repeat
1000 times.
x1 x2 x3 xn x̄ = ni=1 xi
t=1 .21 .76 .57 .84 (.21+.76+. . . )/n
:
:
:
t = 1000
Then, histogram the values in the rightmost column and it looks normal.
1
2
Sampling Distribution of the Sample Variance - Chi-Square Distribution
From the central limit theorem (CLT), we know that the distribution of the sample mean is
approximately normal. What about the sample variance?
Unfortunately there is no CLT analog for variance...

But there is an important special case, which is when X1, X2, . . . , Xn are from a normal
distribution. (Recall that the CLT applies to arbitrary distributions.)
If this is true, the distribution of the sample variance is related to the Chi-Square (χ2) dis-
tribution.
Let Z1, Z2, . . . , Zν be N (0, 1) r.v.’s and let X = Z2 + Z2 + · · · + Z2. Then the pdf of X
1 2 ν
can be shown to be:
1
f (x) = xν/2−1e−x/2 for x ≥ 0.
2ν/2 Γ(ν/2)
This is the χ2 distribution with ν degrees of freedom (ν adjustable quantities). (Note: the
χ2 distribution is a special case of the Gamma distribution with parameters λ = 1/2 and
r = ν/2.)
Fact proved in book:

If X1, X2, . . . , Xn are iid N (µ, σ) r.v.’s, then
(n − 1) 2
S ∼ χ2n−1.
σ2
That is, the sample variance times a constant (n−1) has a χ2 distribution.
σ2 n−1
Technical Note: we lost a degree of freedom when we used the sample mean rather than the
true mean. In other words, fixing n − 1 quantities completely determines s2, since:
1 2
s2 := (xi − x̄) .
n −1 i
Let’s simulate a χ2
n−1
distribution for n = 3. Draw 3 samples from N (0, 1). Repeat 1000
times.
3 2
z1 z2 z3 i=1z i
t=1 -0.3 -1.1 0.2 1.34
:
:
:
t = 1000
Then, histogram the values in the rightmost column.
3
For the chi-square distribution, it turns out that the mean and variance are:
E(χν2) = ν
Var(χν2) = 2ν.
We can use this to get the mean and variance of S2:

� 2 2 �
σ χ σ2
−
E(S2) = E n−1 = (n 1) = σ2 ,
n − 1
� 2 2 � n − 1
σ4 2 σ4 2 2σ4
Var(S2) = Var σn χ −n−1 = Var(χ2n−1 ) = 12(n − 1) = (n − 1).
So we can well estimate S2 when n is large, since − 1) is small
(n Var(S) n −when
1 n is large.
Remember, the χ2 distribution characterizes normal r.v. with known variance. You need to
know σ! Look below, you can’t get the distribution for S2 unless you know σ.
2 (n − 1)S2 2
X1, X2, . . . , Xn ∼ N (µ, σ ) → ∼ χn−1
σ2
4
Student’s t-Distribution
Let X1, X2, . . . , Xn ∼ N (µ, σ2). William Sealy Gosset aka “Student” (1876-1937) was
looking at the distribution of:
X̄ − µ
T = √
S/ n
Contrast T with σ/ √ which we know is N (0, 1).
X̄ −µ
n
So why was Student looking at this?
Because he had a small sample, he didn’t know the variance of the distribution and couldn’t
estimate it well, and he wanted to determine how far x̄ was from µ. We are in the case of:
• N (0, 1) r.v.’s
• comparing X̄ to µ
• unknown variance σ2
• small sample size (otherwise we can estimate σ2 very well by s2.)

Rewrite
X̄ − µ X̄ −µ
σ/√n Z
= _ .
T= √ S2/σ2
S/ n =
√
S2 1
√ √
n σ/ n
The numerator Z is N (0, 1), and the denominator is sort of the square root of a chi-square,
because remember S2(n − 1)/σ2 ∼ χ2 .
n−1
Note that when n is large, S 2/σ2 → 1 so the T-distribution → N (0, 1).
Student showed that the pdf of T is:
. Σ
Γ ν+1 t2 −(ν+1)/2
f (t) = √ 2
1+ −∞<t<∞
πνΓ (ν/2) ν
5
Snedecor’s F-distribution
The F -distribution is usually used for comparing variances from two separate sources.
Consider 2 independent random samples X1, X2, . . . , Xn ∼ N (µ1, σ2) and Y1, Y2, . . . , Yn ∼
1 1 2
N (µ2, σ2). Define S2 and S2 as the sample variances. Recall:
2 1 2
S2(n1 − 1) S2(n2 − 1)
1 2 2 2
2 ∼ χn1−1 and ∼ χn2−1.
σ 1 σ22
The F-distribution considers the ratio:
S2 /σ2 χ2 /(n1 − 1)
1 1
∼ n1−1
.
S2 /σ22
2 χ2n2−1 /(n2 − 1)
When σ2 = σ2, the left hand side reduces to S2/S2.
1 2 1 2
We want to know the distribution of this! Speaking more generally, let U ∼ χν21 and V ∼ χν22 .
Then W = U/ν1 has an F-distribution, W ∼ Fν ,ν .
V/ν2 1 2
The pdf of W is:

. Σ . Σ −(ν1 +ν2 )/2
Γ ((ν1 + ν2)/2) ν1 ν1/2 ν1/2−1 ν1
f (w) = w 1+ w for w ≥ 0.
Γ(ν1/2)Γ(ν2/2) ν2 ν2
There are tables in the appendix of the book which solve:

P (χ2 > χ2 ) = α
ν ν,α
↑ ↑
P (Tν > tν,α) = α

↑ ↑
P (Fν1,ν2 > fν 1,ν 2,α) = α

↑ ↑
Note that because F is a ratio,

1
Fν1,ν2 =
F ν2,ν1
which you might need to use in order to look up the F-scores in a table in the book. Actually,
you will need to know:
1
fν1,ν2,1−α = .
f ν2,ν1,α
6
7
MIT OpenCourseWare
http://ocw.mit.edu
15.075J / ESD.07J Statistical Thinking and Data Analysis

Fall 2011
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

MIT15 075JF11 chpt05

Uploaded by

Copyright:

Available Formats

MIT15 075JF11 chpt05

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MIT15 075JF11 chpt05

Uploaded by

Copyright:

Available Formats

Central Limit Theorem

(Convergence of the sample mean’s distribution to the normal distribution)

Note 2: “converge” means “convergence in distribution:”

Let’s demonstrate the CLT.

Unfortunately there is no CLT analog for variance...

Fact proved in book:

We can use this to get the mean and variance of S2:

• small sample size (otherwise we can estimate σ2 very well by s2.)

The pdf of W is:

There are tables in the appendix of the book which solve:

P (Tν > tν,α) = α

P (Fν1,ν2 > fν 1,ν 2,α) = α

Note that because F is a ratio,

15.075J / ESD.07J Statistical Thinking and Data Analysis

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.