Difference Between Two Proportions

Using the normal approximation,

( )

,
_

2 1
2 1
1 1
1
n n
p p
P P
Z
s s
where,
. proportion population of estimate pooled
2. population in proportion Sample
1. population in proportion Sample
2
1

p
P
P
s
s
X
1
= # of successes in sample 1
X
2
= # of successes in sample 2
n
1
= sample size for group 1
n
2
= sample size for group 2
P
s1
=
1
1
n
X
P
s2
=
2
2
n
X
and
1
]
1

+
+

2 1
2 1
n n
X X
p
We use this test to compare two proportions. We are trying to determine
whether the difference between two proportions is statistically significant or just
sampling error.
(Example 1)
Compare death rates of liver transplants at 2 hospitals. Test at =.05
Hospital A 77/100 died within 6 months
Hospital B 120/200 died within 6 months
2 1 1
2 1 0
:
:
P P
P P


1
-1.96 1.96
.025
.025
Z
60 .
77 .
2
1

s
S
P
P
657 .
300
197
300
120 77

+
P
93 . 2
058 .
17 .
) 200
1
100
1
)( 343 )(. 657 (.
60 . 77 .

+

Z
Reject H
0
P < .05.
2
(Example 2)
Unemployment rates in 2 Counties. Is there a difference in the unemployment
rates? Test at = .05.
County A: 100/400 Unemployed (25%)
County B: 44/200 Unemployed (22%)
2 1 1
2 1 0
:
:
P P
P P


24 .
600
144
22 .
25 .
2
1

p
P
P
s
s
( )( )
81 .
037 .
03 .
200
1
400
1
76 . 24 .
22 . 25 .

,
_

Z
Do not reject H
0
. P > .05.
3
-1.96 1.96
.025
.025
Z
(Example 3)
Is there a difference between the two suppliers in proportion of defectives?
Test at =.05
Suppler A: 24/80 chips =defective
Suppler B: 30/60 chips =defective
2 1 1
2 1 0
:
:
P P
P P


50 .
30 .
2
1

s
S
P
P
386 .
140
54
p
( )( )
41 . 2
083 .
20 .
60
1
80
1
614 . 386 .
50 . 30 .

,
_

Z
Reject H
0
P <. 05.
4
-1.96 1.96
.025
.025
Z
(Example 4)
Effect of estrogen on Alzheimers disease
Test at =.05
Women receiving estrogen: 7/100 developed Alzheimer
Women not receiving estrogen: 27/150 developed Alzheimer
18 .
07 .
2
1

s
S
P
P
2 1 1
2 1 0
:
:
P P
P P


136 .
250
34
p
( )( )
5 . 2
044 .
11 .
150
1
100
1
864 . 136 .
18 . 07 .

,
_

Z
Reject H
0
P < .05.
5
-1.96 1.96
.025
.025
Z
(Example 5)
Direct Mail Should Company use Sweepstakes. Test at =.05
Sweepstakes No
Sweepstakes
Mailed Out 5,000 4,000
# of Orders 100 60
2 1 1
2 1 0
:
:
P P
P P


018 . 9000 / 160
015 .
020 .
2
1

p
P
P
s
S
( )( )
79 . 1
0028 .
005 .
000 , 4
1
000 , 5
1
982 . 018 .
015 . 020 .

,
_

Z
Do not reject H
0
. P > .05.
6
-1.96 1.96
.025
.025
Z
(Example 6)
Who does better under adverse conditions, men or women?
Donner Party: Caught in blizzard and stranded in mountains for 6 months
without food.
Women: 10/34 died
Men: 30/53 died

2 1 1
2 1 0
:
:
P P
P P


46 . 87 / 40
566 .
294 .
2
1

p
P
P
s
S
( )( )
48 . 2
1095 .
272 .
53
1
34
1
54 . 46 .
566 . 294 .

,
_

Z
Reject H
0
. P < .05 The difference in death rates between men and women was
statistically significant. This study was published by Donald K. Grayson in a
journal. His theory was that women have more fat on them (this is necessary so
that the fetus will survive even if there is an insufficient amount of food) than
men. This enables them to do better in times of famine.
7
-1.96 1.96
.025
.025
Z
(Real Example)
Interesting study published in Family Practice: Two groups were compared: One
group of subjects were exposed to the cold by placing their feet in ice water; the other
group (control group) placed their feet in an empty pail. 26/90 in first group
developed cold within five days; 8/90 in control group developed colds. Is difference
statistically significant? (Yes, Z = 3.45 p< .05).
Researchers concluded that mothers are right and not wearing enough clothing in the
winter and thus becoming chilled can lead to colds.
Matched T-Test (Optional Topic)
(non-independent samples, not two groups)
If you do this in MS Excel, this test is known as the
t-Test: Paired Two Sample for Means
(Example 1)
X
1
(Before) X
2
(After) D=(X
1
-X
2
) D
2
150 130 20 400
160 145 15 225
155 150 5 25
170 162 8 64
150 140 10 100
145 125 20 400
170 155 15 225
160 140 20 400
165 163 2 4
160 160 0 0
115 1,843
8
0 :
0) is difference mean ( 0 :
1
0

D
D
H
H

-2.2622
2.2622
.025
.025
t
9
1
) (
/
2 2
9

n
D D n
D
n S
D
t
D
78 . 4
05 . 24
115
9
) 115 ( ) 1843 ( 10
115
2
9

t

Reject H
0
at P<.05. There is a significant difference between the before and after
measures.
9
(Example 2)
X
1
(Before) X
2
(After) D=(X
1
-X
2
) D
2
12 15 -3 9
4 9 -5 25
17 16 1 1
3 9 -6 36
5 5 0 0
18 14 4 16
15 7 8 64
14 8 6 36
8 1 7 49
6 17 -11 121
16 25 -9 81
15 12 3 9
-5 447
0 :
0 :
1
0


D
D

-2.2010
2.2010
.025
.025
t
11
23 .
22
5
11
) 5 ( ) 447 ( 5
5
2
11

t
Do Not Reject H
0
. P >.05. Difference is not significant.
10