You are on page 1of 5

# Lecture Note 9

## MIT 14.30 Spring 2006

Herman Bennett

22 Interval Estimation

## Interval estimation is another approach for estimating a parameter . Interval estimation

consists in nding a random interval that contains the true parameter with probability
(1 ). Such an interval is called condence interval and the probability (1 ) is called
the condence level.

P A(X1 , ..., Xn ) B(X1 , ...Xn ) = 1 (71)

Example 22.1. Assume a random sample from a N (, 2 ) population, where both para
meters are unknown. Find the 90% (symmetric) condence interval of 2 .

Caution: These notes are not necessarily self-explanatory notes. They are to be used as a complement
to (and not as a substitute for) the lectures.

1
Herman Bennett LN9MIT 14.30 Spring 06

Example 22.2. Assume a random sample from a N (, 2 ) population, where the para
meter is unknown and = 2. Find the 95% (symmetric) condence interval of . How

1.96
1.96
The random interval X n , X + n contains the true parameter with 95%

probability. It is wrong to say that lies in the interval with 95% probability... is not a
RV!

23 Useful results

## A RV X is said to have a t-student distribution with parameter v > 0 (degrees of freedom)

if the pdf of X is:

( v+1
2
) 1 1
X t(v) : f (x|v) = v 2 , (72)
x
( 2 ) v (1 + ( v ))(v/2+0.5)

## Let X N (0, 1) and Z 2n be independent RVs. Then, the RV H is distritbuted

t-student with n degrees of freedom.
X
H= t(n) (73)
Z/n

## (See attached graph and table).

Example 23.1. Important result. Assume a normally distributed random sample. Then,

n(X )
S
t(n1) . Prove this result.

23.2 F distribution

A RV X is said to have an F distribution with parameters v1 > 0 and v2 > 0 if the pdf of
X is:
(v1 /2)
( v1 +2v2 ) v1 x(v1 /21)
X F(v1 ,v2 ) : f (x|v1 , v2 ) = v1 , (74)
( 2 )( v22 ) v2 (1 + ( vv12 )x)(v1 /2+v2 /2)

## for 0 < x < and vi positive integer.

3
Herman Bennett LN9MIT 14.30 Spring 06

## Let X 2n and Z 2m be independent RVs. Then, the RV G is distritbuted F with

n and m degrees of freedom.
X/n
G= F(n,m) (75)
Z/m

## 24 Constructing Condence Intervals for

In what follows, we consider 5 possible cases of limited information about the parameter(s) .
For each case we study how to construct condence intervals.

## 24.1 Case 1: N (, V ar()) and V ar() known

We just saw an example of this case in Example 22.2. Note that in this example = ,
= X
, and the V ar() is known since is known.

## 24.2 Case 2: N (, V ar()) and V ar() unknown

Example 24.1. Assume as in Example 22.2 a normal random sample, but now both
and 2 are unknown. Construct a 95% condence interval of .

## 24.3 Case 3: not N ( ) but pmf/pdf known

We just saw an example of this case in Example 22.1. Note that in this example = 2 ,
= S 2 , and the pdf of (a function of in this case) is known and depends only on one
parameter, 2 .

## 24.4 Case 4: not Normal, pmf/pdf unknown, and n > 30

Example 24.2. Assume a random sample of size n from a population f (x), which is
unknown. Construct a 99% condence interval of = E(Xi ).