Lecture 11.

Chapter 11
Estimation: Comparing two populations

11.1 & 11.2 Estimating the difference between

two population means when the population
variances are known and unknown
11.3 Matched pairs experiment
11.4 Estimating the difference between two
population proportions

11.1 Estimating the difference between

two population means (1 – 2) when
12 and 22 are known

• Two independent random samples are drawn from

the two populations of interest.

• Because we are interested in the difference

between two population means, we use the sample
statistic: x1  x 2

1
The sampling distribution of x1  x 2

• x1  x 2 is normally distributed if the (original)

population distributions are normal.

• x1  x 2 is approximately normally distributed if the

(original) populations are not normal, but the sample
sizes are sufficiently large (greater than 30).

• Expected value of x 1  x 2 is 1 – 2
• Variance of x1  x 2 is 12/n1 + 22/n2

The confidence interval of 1 – 2

So, ( X 1  X 2 )  (     )
Z
   

n1 n2

can be considered as a Z-statistic and used to build a

confidence interval for 1 – 2.
The (1 - ) confidence interval estimator of 1 – 2
when the population variances 12 and 22 are
known is:

2 2
(x  x )  z 1  2
1 2   n n
1 2

2
Example 11.1 revised, page 427

A supermarket chain that offers two different types

of discount coupons for its own brand of bread
wants to estimate with 99% confidence the relative
selling power of the two coupons.

The average number of bread loaves sold during a

14-day period using Coupon 1 was 153, and the
average sold during another 14-day period using
Coupon 2 was 142. Assume that the population
standard deviation for both periods was 10 loaves
per day.

Solution
A 99% z-interval estimator for the difference
between two means is:
2 2
(X X ) z 1  2
1 2 a 2 n n
1 2
10 2 10 2
 (153  142)  2.575 
14 14
 11  9.8  [1.2, 20.8].
There are 3 steps to follow:
1. Calculate 𝑥1 = 153 and 𝑥2 = 142
2. Calculate z/2 = z 0.005 = 2.575
3. Calculate the confidence interval
10 2 10 2
(153  142)  2.575   11  9.8  [1.2,20.8].
14 14

3
11.2 Estimating the difference between
two population means (1 – 2) when
the variances 12 and 22 are unknown
• When the population variances 12 and 22 are
unknown, instead of Z - statistic, we use
T-statistic
( X 1  X 2 )  (     )
T
s s

n1 n2

where s12 and s22 are the sample ‘variances’.

• Two cases are considered when producing the
confidence interval:
- Case 1:The population variances are NOT equal.
- Case 2:The two population variances are equal.

Case I: The two variances are unequal

The (1 - ) confidence interval estimate is as follows:

s2 s2
(x  x )  t 1  2
1 2  2 , d.f. n n
1 2
w here 1-  is the confidence lev el and
(s12 n1  s 22 /n2 ) 2
d.f.  2
( s12 n1 ) 2 ( s 22 n2 )

n1  1 n2  1

4
Example 11.3, page 432
Consmers Non-cmrs
• A sample of 30 people was 2560 2008
randomly drawn. Each person 2420
2116
2812
2940
was identified as a consumer or 2364 2828
a non-consumer of high-fibre 2384 2092
2256 2136
cereal for the breakfast. For 2460 3072
each person, the number of 2240 2504
kilojoules consumed at lunch 2540
2492
2480
2356
was recorded. 2944
2260
2744
• Estimate with 95% confidence 2116
2528
the difference between the 3804
average kilojoule intake of 2976
2528
consumers and non-consumers 2372
of high-fibre cereal for lunch. 3388

Solution
Three steps to follow:
1. Calculate x1  2383.2, x 2  2644.4
s1  142.75, s 2  462.61
It appears that the population variances are unequal.
2. Find:
- the degrees of freedom: d.f. = 25.1
- t/2, d.f. = t0.025, 25 = 2.060

3. Calculate the 95% confidence interval:

142.752 462.612
(2383.2  2644.4)  2.060 
10 20
 261.20  232.45  [-493.65,-28.75]

5
 Case II: The two variances are equal

The (1 - ) confidence interval estimate is as follows

1 1
(x  x )  t s 2p   
1 2  2 , d.f .  n1 n2 

where the pooled variance is estimated by:

(n1  1) s12  (n2  1) s22 and d.f. = n1 + n2 - 2

s 2p 
n1  n2  2

Example 11.4, page 435:

• Comparing the durability of original and
replacement shock absorbers: Part 1.
• Two samples of lifetimes (measured in thousands of
km) of type 1 and type 2 shock absorbers are
randomly and independently selected:
– Sample of 8 original shock absorbers shows:
39.6, 34.2, 47.0, 40.9, 50.6, 27.5, 43.5, 36.3
– Sample of 8 replacement shock absorbers shows:
35.7, 52.0, 46.8, 58.5, 45.7, 52.4, 41.3, 43.8
• Estimate with 90% confidence the difference in
mean kilometres travelled until the sock absorbers
fail between the two types of products (assuming
lifetimes of these two types of shock absorbers are
normally distributed).

6
 Solution
Three steps to follow:
1. Calculate x1  39.95, x 2  47.03
s12  54.02, s 22  51.22, and n1  n 2  8.
It appears that the variances are equal.
Calculate the pooled variance
(8  1)(54.02)  (8  1)(51.22)
s 2p   52.62
2. Find: 882
- the degrees of freedom: d.f. = 8 + 8 -2 = 14
- t/2, d.f. = t0.05, 14 = 1.761
3. Calculate the 90% confidence interval:

1 1
(39.95  47.03)  1.761 52.62    7.08  6.39  [-13.47,-0.69]
8 8

11.3 Matched pairs experiment

Example 11.5, page 442
• Comparing the durability of original and
replacement shock absorbers: Part 2.
• To make this comparison without any side effects, eight
cars each had one new original and one new replacement
shock absorbers installed in the rear end and were driven
until the shock absorbers were no longer effective.
• The numbers of km until the shock absorbers fail were
recorded:
Car 1 2 3 4 5 6 7 8
Original s.a X1 42.6 37.2 50 43.9 53.6 32.5 46.5 39.3
Repl. s.a X2 43.8 41.3 49.7 45.7 52.5 36.8 47 40.7
• Find the 95% confidence estimate for the mean
difference between kilometres driven with the original
and replacement shock absorbers.

7
The (1 - ) confidence interval
• Calculate the sample of differences: XD = X1 – X2

Differences XD: -1.2 -4.1 0.3 -1.8 1.1 -4.3 -0.5 -1.4
We assume that XD is normally distributed (Check it by
drawing a histogram). Calculate 𝑋𝐷 , the sample mean, and
SD, the sample standard deviation
• T-statistic
X D  D
T
s D nD
where D = 1 - 2, nD is the size of sample XD.
T-statistic has a t distribution with nD – 1 degrees of
freedom.
• The (1 - ) confidence interval for D = 1 – 2 is
sD
xD  t / 2 ,nD 1
nD

Solution
Three steps to follow:
1. Calculate the values of sample differences

and
x D  - 1.49, s D  1.92
2. Find:
- the degrees of freedom: d.f. = nD – 1 = 7
- t/2, d.f. = t0.025, 7 = 2.365

3. Calculate the 95% confidence interval for D:

1.92
- 1.49  2.365  -1.49  1.61  [-3.10, 0.12]
8

8
11.4 Estimating the difference between
two population proportions (p1 – p2)
• In this section we deal with two populations whose
data are nominal.
• When data are nominal we can (only) ask questions
regarding the proportions of occurrence of certain
outcomes (called successes).
• Thus, we compare the two populations by estimating
the difference between the two population proportions,
p1 – p2.
• Consider statistic (sample proportion) 𝑝1 = 𝑋1 /𝑛1 where
X1 is number of successes in sample of size n1 taken
from the 1st population; and statistic 𝑝2 = 𝑋2 /𝑛2 where
X2 is number of successes in sample of size n2 taken
from the 2nd population.

• The statistic 𝑝1 − 𝑝2 is approximately normally

distributed if n1p1, n1(1 – p1), n2p2, n2(1 – p2) are
all equal to or greater than 5 (since p1, p2 are
unknown, instead of these we use their point
estimators 𝑝1 𝑎𝑛𝑑 𝑝2 ).
• The mean of 𝑝1 − 𝑝2 is p1 – p2.
• The variance of 𝑝1 − 𝑝2 is (p1(1 – p1) /n1)+ (p2(1 –
p2)/n2).
• And, the following statistic may be
considered as Z- statistic:

( pˆ 1  pˆ 2 )  ( p1  p2 )
Z
pˆ 1 (1  pˆ 1 ) pˆ 2 (1  pˆ 2 )

n1 n2

9
Estimating the difference between
two population proportions
• The (1-) confidence interval for the difference
between two population proportions, p1 – p2, is
given by:

p̂1 (1  p̂1 ) p̂ 2 (1  p̂ 2 )
(p̂1  p̂ 2 )  z /2 
n1 n2

• provided n1p1, n1(1 – p1), n2p2, n2(1 – p2) are all

equal to or greater than 5.

Example 11.6, page 448

• A research project employing 22 000 participants was

conducted to estimate, with 95% confidence, the
difference between the proportions of heart attacks
among those who take aspirin and those who do not.

• Half the participants in the research took aspirin and

half took a placebo.

• In a three-year period, 104 of those who took aspirin

and 189 of those who took the placebo had heart
attacks.

10
Solution: We perform the following three steps
1. Calculate
104 189
p̂1   .009455 ; p̂ 2   .01718
11,000 11,000
2. Find z/2 = z0.025 = 1.96
3. The 95% confidence interval for the difference
between the proportion of regular aspirin takers
who had a heart attack and the proportion of
regular placebo takers who had a heart attack is
p̂1 (1  p̂1 ) p̂ 2 (1  p̂ 2 )
(p̂1  p̂ 2 )  z .025 
n1 n2

(.009455)(.999545) (.01718)(.98282)
 (.009455  .01718)  1.96 
11000 11000
 [ .010753, .004697]

Summary: page 454

Home assignment:

- Section 11.1 Exercises pages 430 - 431: 11.3,

11.7, 11.8

- Section 11.2 Exercises pages 438 - 440: 11.12,

11.17, 11.18

- Section 11.3 Exercises pages 444 - 446: 11.26,

11.28

- Section 11.4 Exercises pages 452 - 453: 11.35,

11.39
22

11