STATISTICAL INFERENCE LECTURE

- Dripto Bakshi
Inferential Statistics

 Inferential statistics: the part of statistics that allows

researchers to generalize their findings beyond data
collected.

 Statistical inference: a procedure for making inferences or

generalizations about a larger population from a sample of
that population
Lecture Parts

 Weak Law of Large Numbers & Central Limit Theorem

 Estimation Theory
Weak Law of Large Numbers & Central Limit
Theorem
Pre - Poll / Election Survey
Markov’ Inequality

 Consider any positive random variable X∈ (0, ∞) with density 𝑓𝑋 .

Markov’ Inequality

 Consider any positive random variable X∈ (0, ∞) with density 𝑓𝑋 .

+∞
 E(X) = ‫׬‬0 𝑥 𝑓𝑋 (x) dx
Markov’ Inequality

 Consider any positive random variable X∈ (0, ∞) with density 𝑓𝑋 .

+∞ +∞
 E(X) = ‫׬‬0 𝑥 𝑓𝑋 (x) dx ≥ ‫𝑋𝑓 𝑥 𝑎׬‬ (x) dx , where a > 0
Markov’ Inequality

 Consider any positive random variable X∈ (0, ∞) with density 𝑓𝑋 .

+∞ +∞
 E(X) = ‫׬‬0 𝑥 𝑓𝑋 (x) dx ≥ ‫𝑋𝑓 𝑥 𝑎׬‬ (x) dx , where a > 0

+∞ +∞
 ‫𝑋𝑓 𝑥 𝑎׬‬ (x) dx ≥ ‫𝑎 𝑎׬‬ . 𝑓𝑋 (x) dx = a . P(X ≥ a)
Markov’ Inequality

 Consider any positive random variable X∈ (0, ∞) with density 𝑓𝑋 .

+∞ +∞
 E(X) = ‫׬‬0 𝑥 𝑓𝑋 (x) dx ≥ ‫𝑋𝑓 𝑥 𝑎׬‬ (x) dx , where a > 0

+∞ +∞
 ‫𝑋𝑓 𝑥 𝑎׬‬ (x) dx ≥ a ‫׬‬0 1. 𝑓𝑋 (x) dx = a . P(X ≥ a)

𝐸(𝑋)
 P(X ≥ a) ≤
𝑎
Chebyshev’s Inequality
 Consider any random variable X ~ 𝐹𝑋 (𝜇, 𝜎 2 )
Chebyshev’s Inequality
 Consider any random variable X ~ 𝐹𝑋 (𝜇, 𝜎 2 )

 Define Q = (𝑋 − 𝜇)2 . Q is definitely a positive random variable.

Chebyshev’s Inequality
 Consider any random variable X ~ 𝐹𝑋 (𝜇, 𝜎 2 )

 Define Q = (𝑋 − 𝜇)2 . Q is definitely a positive random variable.

 By Markov’s inequality, for any 𝜀 > 0

𝐸((𝑋 −𝜇)2 ) 𝜎2
P ((𝑋 − 𝜇)2 > 𝜀 2 ) ≤ =
𝜀2 𝜀2
Chebyshev’s Inequality
 Consider any random variable X ~ 𝐹𝑋 (𝜇, 𝜎 2 )

 Define Q = (𝑋 − 𝜇)2 . Q is definitely a positive random variable.

 By Markov’s inequality, for any 𝜀 > 0

2 2 𝐸((𝑋 −𝜇)2 ) 𝜎 2
P ((𝑋 − 𝜇) > 𝜀 ) ≤ = 2
𝜀2 𝜀

𝜎2
 P(|X − µ| ≥ 𝜀) ≤ 𝜀2
Weak Law of Large Numbers

 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

Weak Law of Large Numbers

 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 =
𝑛
Weak Law of Large Numbers

 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 =
𝑛

 What happens if n -> ∞ ?

Convergence in Probability
𝜎2
 E(𝑀𝑛 ) = 𝜇 & V(𝑀𝑛 ) =
𝑛
Convergence in Probability
𝜎2
 E(𝑀𝑛 ) = 𝜇 & V(𝑀𝑛 ) =
𝑛

 According to Chebyshev’s Inequality: For any finite 𝜀 > 0

𝜎2
𝑛 𝜎2
P(| 𝑀𝑛 − µ| ≥ 𝜀) ≤ =
𝜀2 𝑛𝜀 2
Convergence in Probability
𝜎2
 E(𝑀𝑛 ) = 𝜇 & V(𝑀𝑛 ) =
𝑛

 According to Chebyshev’s Inequality: For any finite 𝜀 > 0

𝜎2
𝑛 𝜎2
P(| 𝑀𝑛 − µ| ≥ 𝜀) ≤ =
𝜀2 𝑛𝜀 2

 Hence as n -> ∞ , P(| 𝑀𝑛 − µ| ≥ 𝜀) -> 0 for any 𝜀 > 0

𝑀𝑛 converges in probability to µ
Pre - Poll / Election Survey
The Methodology
 f: The proportion of people respond “voting for Republicans”
The Methodology
 f: The proportion of people respond “voting for Republicans”

 Define a random variable 𝑋𝑗 for any individual j where

𝑋𝑗 = 1 ; “vote for Republican”
= 0 ; “vote for Democrat”
The Methodology
 f: The proportion of people respond “voting for Republicans”

 Define a random variable 𝑋𝑗 for any individual j where

𝑋𝑗 = 1 ; “vote for Republican”
= 0 ; “vote for Democrat”

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 =
𝑛

 𝑀𝑛 denotes the fraction of people responding “voting for Republicans”

The Methodology
 f: The proportion of people respond “voting for Republicans”

 Define a random variable 𝑋𝑗 for any individual j where

𝑋𝑗 = 1 ; “vote for Republican”
= 0 ; “vote for Democrat”

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 = 𝑛

 𝑀𝑛 denotes the fraction of people responding “voting for Republicans”

 Goal: 95% confidence of ≤1% error

P(|𝑀𝑛 − f| ≥ .01) ≤ .05
 𝑋𝑗 ~ Ber (f)
 𝑋𝑗 ~ Ber (f)

 E(𝑋𝑗 ) = f & V(𝑋𝑗 ) = f(1-f)

 𝑋𝑗 ~ Ber (f)

 E(𝑋𝑗 ) = f & V(𝑋𝑗 ) = f(1-f)

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 =
𝑛
 𝑋𝑗 ~ Ber (f)

 E(𝑋𝑗 ) = f & V(𝑋𝑗 ) = f(1-f)

𝑋1 + 𝑋2 + 𝑋3 ….𝑋𝑛
 𝑀𝑛 =
𝑛

𝑓(1−𝑓)
 E(𝑀𝑛 ) = f & V(𝑀𝑛 ) =
𝑛
 Using Chebyshev’s Inequality
V(𝑀𝑛 )
P(|𝑀𝑛 − f| ≥ .01) ≤ 2
(.01)
 Using Chebyshev’s Inequality
𝑓(1−𝑓)
V(𝑀𝑛 ) 𝑛
P(|𝑀𝑛 − f| ≥ .01) ≤ =
(.01)2 (.01)2
 Using Chebyshev’s Inequality
𝑓(1−𝑓)
V(𝑀𝑛 ) 𝑛
P(|𝑀𝑛 − f| ≥ .01) ≤ =
(.01)2 (.01)2

𝑓(1−𝑓)
 We want 𝑛
< 0.05
(.01)2
 𝑓(1−𝑓)
 We want 𝑛
< 0.05
(.01)2

𝑓(1−𝑓)
 Choose the sample size (n) such that 𝑛
< 0.05
(.01)2
 𝑓(1−𝑓)
 We want 𝑛
(.01)2
< 0.05

𝑓(1−𝑓)
 Choose the sample size (n) such that 𝑛
(.01)2
< 0.05

 But we don’t know f

 In fact we are trying to find / estimate f.

 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼

𝑓(1−𝑓)
1
 Therefore 𝑛
≤
(.01)2 4𝑛(.01)2
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼

𝑓(1−𝑓)
1
 Therefore 𝑛
≤
(.01)2 4𝑛(.01)2

1
 Now < 0.05 if n ≥ 50000
4𝑛(.01)2
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼

𝑓(1−𝑓)
1
 Therefore 𝑛
≤
(.01)2 4𝑛(.01)2

1
 Now < 0.05 if n ≥ 50000
4𝑛(.01)2

 So the minimum sample size (n) = 50000 !!!!

Central Limit Theorem
 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

 E(𝑋𝑗 ) = 𝜇 & V(𝑋𝑗 ) =𝜎 2 ∀ 𝑗 = 1,2, … . . 𝑛

Central Limit Theorem
 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

 E(𝑋𝑗 ) = 𝜇 & V(𝑋𝑗 ) =𝜎 2 ∀ 𝑗 = 1,2, … . . 𝑛

 𝑆𝑛 = 𝑋1 + 𝑋2 + 𝑋3 … . 𝑋𝑛
Central Limit Theorem
 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

 E(𝑋𝑗 ) = 𝜇 & V(𝑋𝑗 ) =𝜎 2 ∀ 𝑗 = 1,2, … . . 𝑛

 𝑆𝑛 = 𝑋1 + 𝑋2 + 𝑋3 … . 𝑋𝑛

 E(𝑆𝑛 ) = n𝜇 & V(𝑆𝑛 ) = n 𝜎 2

Central Limit Theorem
 𝑋1 , 𝑋2 , 𝑋3 ,…… 𝑋𝑛 are iid, where 𝑋𝑗 ~ F (𝜇, 𝜎 2 ) ∀ 𝑗 = 1,2, … . . 𝑛

 E(𝑋𝑗 ) = 𝜇 & V(𝑋𝑗 ) =𝜎 2 ∀ 𝑗 = 1,2, … . . 𝑛

 𝑆𝑛 = 𝑋1 + 𝑋2 + 𝑋3 … . 𝑋𝑛

 E(𝑆𝑛 ) = n𝜇 & V(𝑆𝑛 ) = n 𝜎 2

𝑆𝑛 −E(𝑆𝑛 ) 𝑆𝑛 −n𝜇 𝑀𝑛 −𝜇
 Standardize: 𝑍𝑛 = = = 𝜎
𝜎 n
V(𝑆𝑛) n
 E(𝑍𝑛 ) = 0 & V(𝑍𝑛 ) = 1
 E(𝑍𝑛 ) = 0 & V(𝑍𝑛 ) = 1

 Let Z be a standard Normal random variable. i.e Z ~ 𝑁 0,1

 E(𝑍𝑛 ) = 0 & V(𝑍𝑛 ) = 1

 Let Z be a standard Normal random variable. i.e Z ~ 𝑁 0,1

 The Central Limit Theorem states that:

For every c, P(Zn ≤ c) → P(Z ≤ c) as n becomes large enough. (Convergence in

distribution)

P(Z ≤ c) is the standard normal CDF, Φ(c), available from the normal tables
Election problem – (with CLT)
𝑓(1−𝑓)
 E(𝑀𝑛 ) = f & V(𝑀𝑛 ) =
𝑛
Election problem – (with CLT)
𝑓(1−𝑓)
 E(𝑀𝑛 ) = f & V(𝑀𝑛 ) =
𝑛

 Now we want P(|𝑀𝑛 − f| ≥ .01) ≤ .05

Election problem – (with CLT)

𝑓(1−𝑓)
 E(𝑀𝑛 ) = f & V(𝑀𝑛 ) =
𝑛

 Now we want P(|𝑀𝑛 − f| ≥ .01) ≤ .05

𝑀𝑛 −f .01
 P(| |≥ ) ≤ .05
𝑓(1−𝑓) 𝑓(1−𝑓)
𝑛 𝑛
Election problem – (with CLT)

𝑓(1−𝑓)
 E(𝑀𝑛 ) = f & V(𝑀𝑛 ) =
𝑛

 Now we want P(|𝑀𝑛 − f| ≥ .01) ≤ .05

𝑀𝑛 −f .01
 P(| |≥ ) ≤ .05
𝑓(1−𝑓) 𝑓(1−𝑓)
𝑛 𝑛

.01
 P(|Z| ≥ ) ≤ .05
𝑓(1−𝑓)
𝑛
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼ => 𝑓 1 − 𝑓 ≤ ½
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼ => 𝑓 1 − 𝑓 ≤ ½

.01 .01 𝑛
 ≥ 1 = .02 𝑛
𝑓(1−𝑓)
2
𝑛
 f ∈ 0,1 & 𝑓 1 − 𝑓 ≤ ¼ => 𝑓 1 − 𝑓 ≤ ½

.01 .01 𝑛
 ≥ 1 = .02 𝑛
𝑓(1−𝑓)
2
𝑛

.01
 P(|Z| ≥ ) ≤ P(|Z| ≥ .02 𝑛)
𝑓(1−𝑓)
𝑛
 If n = 10000, P(|Z| ≥ .02 𝑛) = P(|Z| ≥ 2) < .05
 If n = 10000, P(|Z| ≥ .02 𝑛) = P(|Z| ≥ 2) < .05

 In fact P(|Z| ≥ .02 𝑛) = .05 implies .02 𝑛 = 1.96 i.e n = 9604.

 If n = 10000, P(|Z| ≥ .02 𝑛) = P(|Z| ≥ 2) < .05

 In fact P(|Z| ≥ .02 𝑛) = .05 implies .02 𝑛 = 1.96 i.e n = 9604.

 And 9604 << 50000 !!!!!

ESTIMATION
Estimation - categories

 Point Estimation

 Interval estimation
Point Estimation
 The Statistic

 The desirable properties of an estimator:

i. Unbiasedness
ii. Consistency
Interval Estimation

• Interval Estimation: an inferential statistical procedure used to

estimate population parameters from sample data through the
building of confidence intervals

• Confidence Intervals: a range of values computed from sample data

that has a known probability of capturing some population parameter
of interest
The Problem Statement

 Find the average height of Indian males.

 But India has 72 Cr males ! It’s impossible to measure everyone.

 Remedy: Collect a sample, say S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }

 Each sample observation is a realisation of a random variable. Why??

 Since the sample is drawn from the populations 𝑋𝑖 ’s are i.i.d.

The Assumption

 Let us assume that height of Indian males are normally distributed

with mean 𝜇 and variance 𝜎 2 .

 Our aim is to find (estimate) 𝜇 and 𝜎 2

 These are called “population parameters”.

Sample Statistics

1
 Sample mean 𝑋 = σ𝑛𝑖=1 𝑋𝑖
ത
𝑛

1
2
 Sample variance 𝑠 = σ𝑛𝑖=1(𝑋𝑖 − 𝑋)
ത 2
𝑛−1
Sample Statistics

1
 Sample mean 𝑋 = σ𝑛𝑖=1 𝑋𝑖
ത
𝑛

1
2
 Sample variance 𝑠 = σ𝑛𝑖=1(𝑋𝑖 − 𝑋)
ത 2
𝑛−1

 I would want to use the sample statistics to “estimate” the

population parameters.

 Let’s see how.

Point Estimation

ത and sample variance (𝑠 2 ) as proxies

 Just take the sample mean (𝑋)
for the population mean (𝜇) and population variance (𝜎 2 ),
respectively.

 𝑋ത and 𝑠 2 are called point estimators of 𝜇 and 𝜎 2 respectively.

 But are they “good” proxies?

The “Good” Properties

The good / desirable properties of a point estimator are

i. Unbiasedness
ii. Consistency
The “Good” Properties
The good / desirable properties of a point estimator are

i. Unbiasedness:
ത =𝜇
 E(𝑋)
 E(𝑠 2 ) = 𝜎 2

ii. Consistency
The “Good” Properties
The good / desirable properties of a point estimator are

i. Unbiasedness:

ത =𝜇
 E(𝑋)
 E(𝑠 2 ) = 𝜎 2

ii. Consistency:

ത =0
 lim 𝑉(𝑋)
𝑛→ ∞
 lim 𝑉(𝑠 2 ) = 0
𝑛→ ∞
Interval estimation of mean
 Sample observations S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }
Interval estimation of mean
 Sample observations S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }

 𝑋𝑖 ~ N (𝜇, 𝜎 2 )
Interval estimation of mean
 Sample observations S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }

 𝑋𝑖 ~ N (𝜇, 𝜎 2 )

1
 𝑋 = σ𝑛𝑖=1 𝑋𝑖
ത
𝑛
Interval estimation of mean
 Sample observations S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }

 𝑋𝑖 ~ N (𝜇, 𝜎 2 )

1
 𝑋 = σ𝑛𝑖=1 𝑋𝑖
ത
𝑛

𝜎2
 𝑋ത ~ N (𝜇, )
𝑛
Interval estimation of mean
 Sample observations S = { 𝑋1 , 𝑋2 , 𝑋3 … . 𝑋𝑛 }

 𝑋𝑖 ~ N (𝜇, 𝜎 2 )

1
 𝑋ത = σ𝑛𝑖=1 𝑋𝑖
𝑛

𝜎2
 𝑋ത ~ N (𝜇, )
𝑛

𝑋ത −𝜇
 Z= 𝜎 ~ N (0,1)
𝑛
The Standard Normal Distribution
The Interval
 P( 𝑍 < 𝑍𝛼/2 ) = 1 - 𝛼
The Interval
 P( 𝑍 < 𝑍𝛼/2 ) = 1 - 𝛼

 P(- 𝑍𝛼/2 < Z < 𝑍𝛼/2 ) = 1 - 𝛼

The Interval
 P( 𝑍 < 𝑍𝛼/2 ) = 1 - 𝛼

 P(- 𝑍𝛼/2 < Z < 𝑍𝛼/2 ) = 1 - 𝛼

𝑋ത −𝜇
 P(- 𝑍𝛼/2 < 𝜎 < 𝑍𝛼/2 ) = 1 - 𝛼
𝑛
The Interval
 P( 𝑍 < 𝑍𝛼/2 ) = 1 - 𝛼

 P(- 𝑍𝛼/2 < Z < 𝑍𝛼/2 ) = 1 - 𝛼

𝑋ത −𝜇
 P(- 𝑍𝛼/2 < 𝜎 < 𝑍𝛼/2 ) = 1 - 𝛼
𝑛

𝜎 𝜎
 P(𝜇 ∈ (𝑋ത − . 𝑍𝛼 , 𝑋ത + . 𝑍𝛼 ) = 1 - 𝛼
𝑛 2 𝑛 2
The Interval
 P( 𝑍 < 𝑍𝛼/2 ) = 1 - 𝛼

 P(- 𝑍𝛼/2 < Z < 𝑍𝛼/2 ) = 1 - 𝛼

𝑋ത −𝜇
 P(- 𝑍𝛼/2 < 𝜎 < 𝑍𝛼/2 ) = 1 - 𝛼
𝑛

𝜎 𝜎
 P(𝜇 ∈ (𝑋ത − . 𝑍𝛼 , 𝑋ത + . 𝑍𝛼 ) = 1 - 𝛼
𝑛 2 𝑛 2

 This is the interval within which 𝜇 lies with probability = 1 - 𝛼

 We can compute 𝑋ത from the sample. n is the sample size. 𝑍𝛼/2 can be computed from
the standard normal table. So if we know 𝜎, we are done !!
We may not know 𝝈 !!
The t - statistic

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2
The t - statistic

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2

𝑋ത −𝜇
 Define a statistic t = 𝑠
𝑛
The t - statistic

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2

ഥ −𝜇
𝑋
𝜎
𝑋ത −𝜇 𝑛
 Define a statistic t = 𝑠 =
𝑠/𝜎
𝑛
The t - statistic

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2

ഥ −𝜇
𝑋 ഥ −𝜇
𝑋
𝜎 𝜎
𝑋ത −𝜇 𝑛 𝑛 𝑍
 Define a statistic t = 𝑠 = = = ~ 𝑡𝑛−1
𝑠/𝜎 𝑛−1 .𝑠2 1
𝑛 .𝑛−1 χ2(𝑛−1)
𝜎2 .
𝑛−1
The t - statistic

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2

ഥ −𝜇
𝑋 ഥ −𝜇
𝑋
𝜎 𝜎
𝑋ത −𝜇 𝑛 𝑛 𝑍
 Define a statistic t = 𝑠 = = = ~ 𝑡𝑛−1
𝑠/𝜎 𝑛−1 .𝑠2 1
𝑛 .𝑛−1 χ2(𝑛−1)
𝜎2 .
𝑛−1

i.e the student’s t – distribution with n-1 degrees of freedom.

Student’s t - distribution
The Interval
(𝛼/2)
 P( 𝑡 < 𝑡𝑛−1 )=1-𝛼
The Interval
(𝛼/2)
 P( 𝑡 < 𝑡𝑛−1 )=1-𝛼

(𝛼/2) (𝛼/2)
 P(-𝑡𝑛−1 < t< 𝑡𝑛−1 ) =1-𝛼

(𝛼/2) 𝑋ത −𝜇 (𝛼/2)
 P(-𝑡𝑛−1 < 𝑠 <𝑡𝑛−1 ) =1-𝛼
𝑛
The Interval
(𝛼/2)
 P( 𝑡 < 𝑡𝑛−1 )=1-𝛼

(𝛼/2) (𝛼/2)
 P(-𝑡𝑛−1 < t < 𝑡𝑛−1 ) = 1 - 𝛼

(𝛼/2) 𝑋ത −𝜇 (𝛼/2)
 P(-𝑡𝑛−1 < 𝑠 <𝑡𝑛−1 ) =1-𝛼
𝑛

𝑠 (𝛼/2) ത 𝑠 (𝛼/2)
 P(𝜇 ∈ (𝑋ത − . 𝑡𝑛−1 , 𝑋 + . 𝑡𝑛−1 ) =1-𝛼
𝑛 𝑛
The Interval
(𝛼/2)
 P( 𝑡 < 𝑡𝑛−1 ) = 1 - 𝛼

(𝛼/2) (𝛼/2)
 P(-𝑡𝑛−1 < t < 𝑡𝑛−1 ) = 1 - 𝛼

(𝛼/2) 𝑋ത −𝜇 (𝛼/2)
 P(-𝑡𝑛−1 < 𝑠 <𝑡𝑛−1 ) = 1 - 𝛼
𝑛

𝑠 (𝛼/2) 𝑠 (𝛼/2)
 P(𝜇 ∈ (𝑋ത − . 𝑡𝑛−1 , 𝑋ത + . 𝑡𝑛−1 ) = 1 - 𝛼
𝑛 𝑛

 This is the interval within which 𝜇 lies with probability = (1 - 𝛼)

(𝛼/2)
 We can compute 𝑋ത from the sample. n is the sample size.𝑡𝑛−1 can be computed from the t - distribution
table.
Interval estimation of Variance

𝑛−1 .𝑠 2
 ~ χ2(𝑛−1)
𝜎2
 𝛼 𝛼
2
(1− ) 𝑛−1 .𝑠2 ( )
2 2
 P( χ 𝑛−12 < < χ 𝑛−1 ) = 1-𝛼
𝜎2

2 𝑛−1 .𝑠2 𝑛−1 .𝑠2

 P (𝜎 ∈ ( 𝛼
(1− 2 )
, 𝛼 )=1-𝛼
χ 2
𝑛−1 χ 2 2
𝑛−1

 Given a sample we can compute 𝑠 2 and we know n (sample size). We can

𝛼 𝛼
( ) (1− )
2 2 2
compute χ 𝑛−1 & χ 𝑛−12 from the chi – square distribution table.

 Thus we have found the interval such that 𝜎 2 lies in that interval with probability
(1- 𝛼 )