Interval Estimation

Although point estimators are useful, interval estimators convey more information about
the data. The purpose of using an interval estimator is to have some degree of confidence of
securing the true parameter. Interval estimators are called “Confidence Intervals” and the limits
are called the upper and lower “Confidence Limits”.
Let L and U be the lower and upper confidence limits for a parameter 𝜃 based on a random
sample 𝑋1 , 𝑋2 , … , 𝑋𝑛 ,
𝑃(𝐿 ≤ 𝜃 ≤ 𝑈) = 1 − 𝛼
We are (1 − 𝛼)% confident that the true parameter 𝜃 is located in the interval (𝐿, 𝑈).
Confidence Interval for mean-when 𝜎 2 is known
2
In a random sample 𝑋1 , 𝑋2 , … , 𝑋𝑛 from 𝑁(𝜇, 𝜎 2 ), 𝑋̅~𝑁(𝜇, 𝜎 ⁄𝑛). By standardizing 𝑋̅,

𝑃(−𝑍𝛼⁄2 ≤ 𝑍 ≤ 𝑍𝛼⁄2 ) = 1 − 𝛼

𝑋̅ − 𝜇
𝑃 (−𝑍𝛼⁄2 ≤ 𝜎 ≤ 𝑍𝛼⁄2 ) = 1 − 𝛼
⁄ 𝑛
√
𝜎 𝜎
𝑃 (𝑋̅ − 𝑍𝛼⁄2 ≤ 𝜇 ≤ 𝑋̅ + 𝑍𝛼⁄2 ) = 1 − 𝛼
√𝑛 √𝑛
𝜎 𝜎
Therefore, a (1 − 𝛼)% confidence interval for 𝜇 is [𝑋̅ − 𝑍𝛼⁄2 , 𝑋̅ + 𝑍𝛼⁄2 ]
√𝑛 √𝑛

Interpretation
If samples of the same size n are drawn repeatedly from a population, and a confidence
interval is calculated from each sample, then (1 − 𝛼)% of these intervals should contain the
population mean. This is often stated as,
𝜎 𝜎
We are (1 − 𝛼)% confident that the true mean is in the interval [𝑋̅ − 𝑍𝛼⁄2 , 𝑋̅ + 𝑍𝛼⁄2 ]
√𝑛 √𝑛

Example:
A random sample of 12 item is selected from a normal population with variance 4. Measurements
of some characteristics of the sample is as follows.
9.5 9.5 11.2 10.6 9.9 11.1 10.9 9.8 10.1 10.2 10.9 11.0
Calculate a 95% confidence interval for the population mean 𝜇.
Confidence Interval for mean-when 𝜎 2 is unknown
Let 𝑋1 , 𝑋2 , … , 𝑋𝑛 be a random sample of size 𝑛 from 𝑁(𝜇, 𝜎 2 ), where 𝜎 2 is unknown.
𝑋̅ − 𝜇
𝑇= ~𝑡𝛼⁄2,𝑛−1
𝑆⁄
√𝑛
Then, 𝑃(−𝑡𝛼⁄2,(𝑛−1) ≤ 𝑇 ≤ 𝑡𝛼⁄2,(𝑛−1) ) = 1 − 𝛼

𝑋̅ − 𝜇
𝑃 (−𝑡𝛼⁄2,(𝑛−1) ≤ ≤ 𝑡𝛼⁄2,(𝑛−1) ) = 1 − 𝛼
𝑆⁄
√𝑛
𝑆 𝑆
𝑃 (𝑋̅ − 𝑡𝛼⁄2,(𝑛−1) ≤ 𝜇 ≤ 𝑋̅ + 𝑡𝛼⁄2,(𝑛−1) ) = 1 − 𝛼
√𝑛 √𝑛
𝑆
Therefore (1 − 𝛼)% confidence interval for 𝜇 when 𝜎 2 is unknown is, [𝑋̅ ± 𝑡 ⁄ ].
√𝑛 𝛼 2,(𝑛−1)

Example:
10 packets of particular brand of biscuits are selected at random and their weights are noted.
397.3 399.6 401 392.9 396.8 400 397.6 392.1 400.8 400.6
Assuming that the sample is taken from a normal population with mean weight 𝜇 and calculate
99% confidence interval for 𝜇.

Confidence Interval for Proportion

𝑌
Let 𝑌~𝐵𝑖𝑛(𝑛, 𝑝) and a point estimator of the parameter p is given by the statistics 𝑝̂ = 𝑛;
Where 𝑌 denote the number of successes in 𝑛 trials.
𝑝(1−𝑝)
𝐸(𝑝̂ ) = 𝑝 𝑉(𝑝̂ ) = 𝑛
𝑝(1−𝑝)
Therefore, for large 𝑛, 𝑝̂ ~𝑁(𝑝, ),
𝑛

𝑝̂ − 𝑝
𝑍= ~𝑁(0,1)
√𝑝(1 − 𝑝)⁄𝑛

(1 − 𝛼)% confidence interval for 𝑝,

𝑃(−𝑍𝛼⁄2 ≤ 𝑍 ≤ 𝑍𝛼⁄2 ) = 1 − 𝛼
 𝑝̂ − 𝑝
𝑃 −𝑍𝛼⁄2 ≤ ≤ 𝑍𝛼⁄2 = 1−𝛼
√𝑝(1 − 𝑝)⁄𝑛
( )

𝑝(1 − 𝑝) 𝑝(1 − 𝑝)
𝑃 (𝑝̂ − 𝑍𝛼⁄2 √ ≤ 𝑝 ≤ 𝑝̂ + 𝑍𝛼⁄2 √ )=1−𝛼
𝑛 𝑛

𝑝(1−𝑝) 𝑝̂(1−𝑝̂)
When 𝑛 is large, we can estimate the standard deviation √ by √ , because 𝑝̂ is an
𝑛 𝑛
unbiased estimator for population proportion 𝑝. Then,

𝑝̂ (1 − 𝑝̂ ) 𝑝̂ (1 − 𝑝̂ )
𝑃 (𝑝̂ − 𝑍𝛼⁄2 √ ≤ 𝑝 ≤ 𝑝̂ + 𝑍𝛼⁄2 √ )=1−𝛼
𝑛 𝑛

Example:
A manufacture wants to assess the proportion of defective items in a large batch produced by a
particular machine. A random sample of 300 items is taken and it is observed that 45 items of them
are defective. Calculate a 95% confidence for the proportion of defective items in the complete
batch.

Confidence Interval for Variance

Let 𝑋1 , 𝑋2 , … , 𝑋𝑛 be a random sample of size 𝑛 from 𝑁(𝜇, 𝜎 2 ), where mean 𝜇 is unknown.
Then,

2
(𝑛 − 1)𝑆 2 2
𝜒 = ~𝜒𝑛−1
𝜎2
We can determine values a and b from the chi-square distribution such that,
(𝑛 − 1)𝑆 2
𝑃 (𝑎 < < 𝑏) = 1 − 𝛼
𝜎2
2 2
Here 𝑎 = 𝜒1−𝛼 ⁄2,𝑛−1 and 𝑏 = 𝜒𝛼⁄2,𝑛−1 . Then,

(𝑛 − 1)𝑆 2 2
(𝑛 − 1)𝑆 2
𝑃( 2 ≤𝜎 ≤ 2 )= 1−𝛼
𝜒𝛼⁄2,𝑛−1 𝜒1−𝛼⁄2,𝑛−1

Example:
The followings are the weights of 10 package of grass seed distributed by a certain company.
46.4 46.1 45.8 47 46.1 45.9 45.8 46.9 45.2 46
Find 95% confidence for the variance of all such packages of grass seed distributed by this
company, assuming a normal population.

Confidence Interval for the mean difference

𝝈𝟐𝟏 𝒂𝒏𝒅 𝝈𝟐𝟐 known.
Let 𝑋1 , 𝑋2 , … , 𝑋𝑛 and 𝑌1 , 𝑌2 , … , 𝑌𝑚 be independent random sample of size n and m from
two independent normal populations with mean 𝜇1 , 𝜇2 and variances 𝜎12 , 𝜎22 respectively.
2 2
𝜎 𝜎
𝑋̅~𝑁(𝜇1 , 1 ⁄𝑛) 𝑌̅~𝑁(𝜇2 , 2 ⁄𝑚)
𝜎12⁄ 𝜎22⁄ (𝑋̅ −𝑌̅)−(𝜇1 −𝜇2 )
Then (𝑋̅ − 𝑌̅)~𝑁 (𝜇1 − 𝜇2 , 𝑛 + 𝑚) and 𝑍= 2 2
√𝜎1 ⁄ +𝜎2 ⁄
𝑛 𝑚

Therefore, (1 − 𝛼)% confidence interval for 𝜇1 − 𝜇2 ,

𝑃(−𝑍𝛼⁄2 ≤ 𝑍 ≤ 𝑍𝛼⁄2 ) = 1 − 𝛼

(𝑋̅ − 𝑌̅) − (𝜇1 − 𝜇2 )

𝑃 −𝑍𝛼⁄2 ≤ ≤ 𝑍𝛼⁄2 = 1−𝛼
2 2
√𝜎1 ⁄ + 𝜎2 ⁄
( 𝑛 𝑚 )

2 2 2 2
𝜎 𝜎 𝜎 𝜎
𝑃 ((𝑋̅ − 𝑌̅) − 𝑍𝛼⁄2 √ 1 ⁄𝑛 + 2 ⁄𝑚 ≤ (𝜇1 − 𝜇2 ) ≤ (𝑋̅ − 𝑌̅) + 𝑍𝛼⁄2 √ 1 ⁄𝑛 + 2 ⁄𝑚) = 1 − 𝛼

𝝈𝟐𝟏 = 𝝈𝟐𝟐 = 𝝈𝟐 but 𝝈𝟐 is unknown

(𝑋̅ − 𝑌̅) − (𝜇1 − 𝜇2 )
𝑇= ~𝑡𝛼⁄2,𝑚+𝑛−2
1 1
𝑆𝑝 √𝑛 + 𝑚

(𝑛−1)𝑆12 +(𝑚−1)𝑆22
Where 𝑆𝑝2 = , then (1 − 𝛼)% confidence interval for 𝜇1 − 𝜇2 ,
𝑚+𝑛−2

1 1
(𝑋̅ − 𝑌̅) ± 𝑡𝛼⁄2,𝑚+𝑛−2 𝑆𝑝 √ +
𝑛 𝑚

𝝈𝟐𝟏 ≠ 𝝈𝟐𝟐 and unknown

(1 − 𝛼)% confidence interval for 𝜇1 − 𝜇2 ,

𝑆12 𝑆22
(𝑋̅ − 𝑌̅) ± 𝑡𝑑 ,𝛼⁄2 √ +
𝑛 𝑚
 𝑆 2
𝑆 2 2
( 1+ 2)
𝑛 𝑚
where, Degrees of freedom = 𝑑 = 2 2
𝑆2 𝑆2
( 1) ( 2)
𝑛 𝑚
+
𝑛−1 𝑚−1

Example:
From a large class 4 students were drawn and their maths marks were 64,66,89,77. From a second
large class 3 students were drawn and their maths marks were 56,71,53. Assuming the classes are
normally distributed and have equal variances, find 95% confidence interval for the difference
between the mean of the two maths marks.

Confidence interval for the difference between two Proportions

𝑝1 (1−𝑝1 ) 𝑝2 (1−𝑝2 )
Let 𝑝̂1 ~𝑁(𝑝1 , ) and 𝑝̂2 ~𝑁(𝑝2 , )
𝑛1 𝑛2

𝑝1 (1−𝑝1 ) 𝑝2 (1−𝑝2 )
Then, (𝑝̂1 − 𝑝̂ 2 )~𝑁(𝑝1 − 𝑝2 , + )
𝑛1 𝑛2

(𝑝̂1 −𝑝̂2 )−(𝑝1 −𝑝2 )

By standardizing, 𝑍 = 𝑝 (1−𝑝1 ) 𝑝2 (1−𝑝2 )
~𝑁(0,1)
√ 1 +
𝑛1 𝑛2

Therefore, (1 − 𝛼)% confidence interval for the difference of two binomial parameters 𝑝1 − 𝑝2
is,

𝑝̂1 (1 − 𝑝̂1 ) 𝑝̂ 2 (1 − 𝑝̂ 2 )
[(𝑝̂1 − 𝑝̂2 ) ± 𝑍𝛼⁄2 √ + ]
𝑛1 𝑛2

Example:
Two machines are used to produce the same good. In 400 articles produced by machine A, 16 were
substandard. In the same length of time, second machine produced 600 articles, and 60 were
substandard. Construct 95% confidence interval for the difference between two proportions.

Confidence Interval for ratio of two Variances

Let 𝑋~𝑁(𝜇1 , 𝜎12 ) and 𝑌~𝑁(𝜇2 , 𝜎22 )
Then we can write,
(𝑛−1)𝑆12 2 (𝑚−1)𝑆22 2
~𝜒𝑛−1 and ~𝜒𝑚−1
𝜎12 𝜎22

If we divide a chi-square distribution as follows, then we will get a F-distribution with related
degrees of freedom,
 𝜒𝑛2⁄
𝑛
𝐹= 2 ~𝐹𝑛,𝑚
𝜒𝑚⁄
𝑚
Using above result we get that,
𝑆12 𝜎22
𝐹 = 2 2 ~𝐹𝑛−1,𝑚−1
𝑆2 𝜎1
𝑆 2 𝜎2
Then, 𝑃 (𝐹1−𝛼⁄2,𝑛−1,𝑚−1 ≤ 𝑆12 𝜎22 ≤ 𝐹𝛼⁄2,𝑛−1,𝑚−1 ) = 1 − 𝛼
2 1

𝑆22 𝜎22 𝑆22

𝑃 (𝐹1−𝛼⁄2,𝑛−1,𝑚−1 ≤ ≤ 𝐹𝛼⁄2,𝑛−1,𝑚−1 2 ) = 1 − 𝛼
𝑆12 𝜎12 𝑆1

Example:
Two independent random samples of water bottles were drawn from two automatic bottle filling
machines in a factory. Volume of water in each selected bottle was measured 25 bottles from
machine 1 had 𝑆12 = 1.63 and 16 observations from machine 2 had 𝑆22 = 1.23. Construct 90%
𝜎2
confidence interval for the ratio 𝜎22 .
1