7
Checking conditions for CI: random sample,
np
$
560 10
and
n p
(
$
) ( . )1 1000 1 0 56 440 10
Conditions are satisfied. We use :
$
*
$
(
$
)
p z
p p
n
±
−
1
Thus, using the formula above (with z* = 1.96), or using the A:1-PropZInt menu on the calculator, we get
(0.529, 0.591).
That is, based on the results from our sample of size 1000, we are 95% confident that the
proportion of ALL voters who favored Propostion 1 is between 52.9% and 59.1%.
Notice that the sample size of 1000 gives a much narrower confidence interval than the sample
size of 100. In fact, with the larger sample, we can be quite confident (about 95% of the time
anyway), that a majority of the voters favored Proposition 1, since the smaller endpoint of the
samples 95% confidence interval, 0.529 is greater than one-half. Bear in mind, however, that the
larger sample may be more costly and time consuming than the smaller one.
Now, how confident are you that Proposition 1 passed or failed?
I’d bet a small amount of money that I am right.
c. Forget the previous parts now. Assume that you didn’t take any samples yet. What sample size
you need to use if you want the margin of error to be at most 3% with 95% confidence but you
have no estimate of p?
Because you don’t have an estimate of p, use
$
p
= 0.5. We want the margin of error to be at most
3%, that is m = 0.03.
n
z
m
p p
=
− =
− =
*
$
(
$
)
.
.
. ( . ) .
2
2
1
196
0 03
0 5 1 0 5 1067111
Thus, to get a margin of error to be at most 3%, we need at least 1068voters in our sample.
d. Now let’s assume you did a pilot sample, in which 56 out of 100 voters said they favor
Proposition 1. What sample size you need to use if you want the margin of error to be at most
3% with 95% confidence now?
Now we have an estimate of p from the pilot study, so we use
$
p
= 0.56. We want the margin of
error to be at most 3%, that is m = 0.03.
n
z
m
p p
=
− =
− =
*
$
(
$
)
.
.
. ( . ) .
2
2
1
196
0 03
056 1 0 56 105174
Thus, to get a margin of error to be at most 3%, we need at least 1052 voters in our sample.
6. Sometimes a 95% confidence interval is not enough. For example, in testing new medical drugs
or procedures, a 99% confidence interval may be required before the new drug or procedure is
approved for general use. For example, a new drug for migraines might induce insomnia
(difficulty of falling asleep) in some patients. If this side effect happens in too many patients, the