Transcript week4_sampling

2007 Fall_COMM 420_Week 4(1) @ NY
Research Methods in AD/PR
COMM 420
Section 8
Tuesday / Thursday 3:35 pm -5:30 pm
143 Stuckeman
Nan Yu
1
2007 Fall_COMM 420_Week 4(1) @ NY
Population


“Parameter”
Census


If you question
every member
of the
population
Universe

Words, news,
characters
2007 Fall_COMM 420_Week 4(1) @ NY
Sample
Can a small group of people represent a larger population?
Yes, but we need to make the sample representative.
2007 Fall_COMM 420_Week 4(1) @ NY
Sample
A sample is a representative group of
people similar to the population.
2007 Fall_COMM 420_Week 4(1) @ NY
Sampling


Sampling: the process of choosing your
sample
Goal


To ensure the sample are representative of
the target population.
To reduce selection bias
2007 Fall_COMM 420_Week 4(1) @ NY
Two types of sampling

Probability sampling

Non-probability sampling
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sampling (ideal)

Requirements



Every person in the population should have
an equal chance of being chosen.
No one will be excluded due to any
reasons.
Every one in the population has a specific
and known probability of being included in
your sample.
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sampling
What is the probability
of selecting a ball in
the box?
1/10=10%
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sampling
If a red ball is selected
and taken out of the box,
what is the probability of
selecting
another ball in the box?
1/9=11.1%
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sampling
If another blue ball is
selected
and taken out of the box,
what is the probability of
selecting a ball in the box?
1/8=12.5%
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sampling
As we select more and more
balls and take them out of the
boxes, the probability of
selecting a ball has increased.
1/10=10%
1/9=11.1%
1/8=12.5%
2007 Fall_COMM 420_Week 4(1) @ NY
Against the rule!
Probability Sample —

Every one in the population has a specific
and known probability of being included in
your sample.
2007 Fall_COMM 420_Week 4(1) @ NY
So, what we should do is…
We take one ball out, put
it back, mix them up,
then, draw another ball…
So the probability of
selecting one ball is
always 1/10=10%
2007 Fall_COMM 420_Week 4(1) @ NY
Probability Sample
Every person in the population should have an
equal chance of being chosen.
2007 Fall_COMM 420_Week 4(1) @ NY
Types of probability sampling

Simple random sampling
An ideal situation
that researchers
always try to
achieve
With this method, each member of
the population has a statistically
equal chance of being selected as a
sample, thus reducing bias in the
sample.
2007 Fall_COMM 420_Week 4(1) @ NY
Types of probability sampling

Stratified Random Sampling
•Very often used in political opinion
polls, they will break down the
population first by
•Sex
•Age
•Race
•……
Then random select people from
each group.
2007 Fall_COMM 420_Week 4(1) @ NY
Types of probability sampling

Systematic Random Sampling
•#2, #10 on each page of a
telephone book
•opinion page of the New York
Times on every Mondays and
Thursdays from 2003-2007
2007 Fall_COMM 420_Week 4(1) @ NY
Non-probability sample


Not everyone in the population has an
equal chance to be chosen.
We choose people that we think match
the population characteristics.
2007 Fall_COMM 420_Week 4(1) @ NY
Types

Convenience sample




Limitations of times/resources.
Sampling whoever you can get
conveniently.
This approach is commonly used in the
academic-orientated studies, but not in the
real-world research.
Location biases, time biases,…etc.
2007 Fall_COMM 420_Week 4(1) @ NY
Types

Quota sampling



The population is first segmented into
mutually exclusive sub-groups
Then choose subjects or units from each
segment based on a specified proportion.
If in a population, 80% are female, 20%
are male. You need to make sure that the
sample that you create follow the similar
proportion.
2007 Fall_COMM 420_Week 4(1) @ NY
Volunteer Sampling

Participants are rewarded in some way.



Class credit
Money…
Motivation biased, location biased…
2007 Fall_COMM 420_Week 4(1) @ NY
How to contact your
participations





Telephone
Mail
Face-to-face, door-to-door
Online questionnaires
Computerized telephone
2007 Fall_COMM 420_Week 4(1) @ NY
Sampling size and error

Sampling error may introduced by




the process itself
biases
It may be reduced by increasing the
sample size.
By reducing the sampling error, we are
trying to make the sample as
representative as possible.
2007 Fall_COMM 420_Week 4(1) @ NY
Video Time


Sampling and Estimation
Deadly Deception