### Presentation Title

```Statistical Sampling
Part III – Confidence Intervals
This video is designed to accompany
pages 41-76
in
Making Sense of Uncertainty
Activities for Teaching Statistical Reasoning
Van-Griner Publishing Company
Error Due to Sampling
Americans and Their Guns
Title:
Poll: Majority of Americans Back Stricter Gun Laws
Authors: Sarah Dutton, Jennifer De Pinto, Anthony Salvanto, Fred Backus, Leigh Ann
Caldwell
Source: CBS News January 17, 2013
As the president outlined sweeping new proposals aimed to reduce gun violence, a new
CBS News/New York Times poll found that Americans back the central components of
the president's proposals, including background checks, a national gun sale database,
limits on high capacity magazines and a ban on semi-automatic weapons. Asked if they
generally back stricter gun laws, more than half of respondents - 54 percent - support
stricter gun laws …. That is a jump from April - before the Newtown and Aurora
shootings - when only 39 percent backed stricter gun laws but about the same as ten
years ago.
…
This poll was conducted by telephone from January 11-15, 2013 among 1,110 adults
nationwide. Phone numbers were dialed from samples of both standard land-line and
cell phones. The error due to sampling for results based on the entire sample could be
plus or minus three percentage points.
Margin of Error
The 3% is called the “margin of error” for the
survey.
•
•
You may never have heard about it before now.
•
Our goal is to make sure you understand what it really
is and isn’t.
Fundamental Problem
Another, equally well-chosen SRS of n = 1,110
almost surely yield a different p.
And it can be quite different from the 0.54 seen
here.
Sampling Variability
The variability seen in a statistic from sample to
sample is called “sampling variability.”
Question:
Which statistic is correct?
They are all correct!
Better
How can you estimate the
parameter in
Question:
the face of this variability?
Mathematics to the Rescue
With SRS-type samples, sampling variability is
understandable, predictable.
With convenience samples it is not.
If sample is not a probabilistic sample, then it will
be very difficult to do the formal inference with
integrity.
Predictable? How?
1. If you were to do the sampling over and over and
plot the different statistics you get ….
2. Then that plot – called a sampling distribution –
would exhibit predictable characteristics.
3. In particular, it would be bell-shaped and peak
above the parameter from the population.
Predictable? How?
Suppose you take a SRS of size 80 and ask “Are you in favor of samesex marriages?” Record the proportion who say “Yes.” Put those 80
back in the population and take another sample of size 80. Do this 25
times.
Number of Sample
Proportions Observed
A Sampling Distribution
7
6
5
4
3
2
1
0
.41 to .46 to .51 to .56 to .61 to .66 to .71 to .76 to .81 to .86 to .91 to
.45
.50
.55
.60
.65
.70
.75
.80
.85
.90
.95
Interval for Sample Proportion
A histogram of these 25
sample proportions would
have to be bell-shaped
and peak above the
parameter in the
population, which appears
to be about 2/3 in this
case.
Enormously Useful
Can make quantitative statements
about how far p is likely to be
from p, even if you don’t know p.
From the graphic, one
can see that 95% of
p’s based on samples
of size n, will be within
1
of p.
n
95% Confidence Interval
If you are estimating a proportion p …
From a simple random sample of size n …
And want 95% “confidence” in your estimate …
Then your margin of error is MOE =
1
n
And a 95% confidence interval for p is
1
p +/−
n
…
Example Revisited
Americans and Their Guns
Title:
Poll: Majority of Americans Back Stricter Gun Laws
Authors: Sarah Dutton, Jennifer De Pinto, Anthony Salvanto, Fred Backus, Leigh Ann
A 95% confidence interval for p is
Caldwell
1
Source: CBS News January 17, 2013
0.54 +/−
1110
=0.54 +/- 0.03
As the president outlined sweeping new proposals aimed to reduce gun violence, a new
CBS News/New York Times poll found that Americans back the central components of
the president's proposals, including background checks, a national gun sale database,
limits on high capacity magazines and a ban on semi-automatic weapons. Asked if they
generally back stricter gun laws, more than half of respondents - 54 percent - support
stricter gun laws …. That is a jump from April - before the Newtown and Aurora
shootings - when only 39 percent backed stricter gun laws but about the same as ten
years ago.
…
This poll was conducted by telephone from January 11-15, 2013 among 1,110 adults
nationwide. Phone numbers were dialed from samples of both standard land-line and
cell phones. The error due to sampling for results based on the entire sample could be
plus or minus three percentage points.
Changing Confidence?
∗ 1
p +/−
2 n
Level of Confidence
Z*
Level of Confidence
Z*
50%
0.67
90%
1.64
60%
0.84
95%
2.00
70%
1.04
99%
2.58
80%
1.28
99.9%
3.29
No Probabilistic Sample?
∗ 1
p +/−
2 n
Numbers can still be produced
from this type of formula since
plugging in is easy! But if the
samples were not SRS-like,
then those numbers are
meaningless.
A Technical Interpretation
This kind of plot describes
where p is likely to fall, relative
to p, for any SRS of size n.
Since 95% of p’s based on
samples of size n, will be
1
within
of p, it follows that
n
for 95% of all the p’s
observed, the range  +/−
1
n
will be wide enough to cover p.
Illustration of Interpretation
Parameter
p = 0.79
95% confidence” means to 95%
of a long list of these intervals
would contain the parameter!
0.60
0.70
0.80
0.90
One-Sentence Reflection
Simple formulas are available for the
margin of error and associated
confidence intervals, provided the data
were collected in a simple random
sample, or similarly statistically correct
fashion.
```