This is a five-section course as part of a two-course sequence in Research Methods in Psychology. This course deals with descriptive methods and the second course deals with experimental methods.
Lesson 4A – Populations and Samples
Loading...
來自 Georgia Institute of Technology 的課程
Descriptive Research Methods in Psychology
2 個評分
從本節課中
Module 4: Sampling
與講師見面
Dr. Anderson D. SmithRegents’ Professor Emeritus
School of Psychology
[MUSIC]
Anderson Smith again, and today we're going to talk about sampling.
Whether you're watching chimpanzees in the wild like Jane Goodall, studying
a classroom and looking at children through a one way observation mirror, or
whether you're doing a survey in a shopping mall, you're trying to represent
some particular population by doing your research on a sample of that population.
So, a sample is a subgroup of people from a larger population,
that you collect data and analyze and you want to make inferences about
the population from the sample that you're using.
So we have a population of people, it could be large, it could be small,
it's just what you're trying to study, the group you're trying to study.
And because you can't study every member of that population, you take a sample.
And that sample then is assumed to represent
the population you want to study.
You make the analysis on the data that you collect in that sample, and
then that you make inferences back to the population at large.
And so this process of what you sample and
how you sample becomes a very important one in descriptive statistics.
So the accuracy of the inferences you make will depend upon how representative
the sample is of the population.
And the question you're trying to answer, what type of sampling do I use?
And what is the sample size, how many people do I have to have in my sample to
really be representative of the population?
And what analyses do I do?
And what analyses am I allowed to do,
given the nature of the data that I'm collecting?
So what are the types of sampling?
Well, the best sampling would be probability sampling,
a representative sampling, where you try ensuring that you
are making a representation that's good of the population that you're trying to test.
And the two kinds of probability sampling are random sampling and
then stratified sampling.
We'll talk about those in more detail.
But you also sometimes have to do nonprobabilty sample,
which is really not representative of the population you're trying to use.
So there's some things you can do to make those better.
You can use quota sampling,
if you have some idea of what types of people you want to have in your sample.
You can use purposive sampling,
where you actually are targeting a particular subgroup.
Or you could do convenience sampling which, by the way,
is what most research actually use and we'll talk about that in some detail.
So we have different kinds of sampling methods.
We have probability sampling, we have non-probability sampling,
random and stratified, quota, purposive and convenience.
Now let's talk about each one of those individually.
First, the two kinds of probability sampling.
And remember, probability samples are actually selected to be representative of
the population, so these are the best to use.
And there are two types, random and stratified.
Now random samples are where each individual
in the population that you're interested in has an equal chance of being selected.
So you have the total population that you're trying to study and
you have an equal probability of every person that you select
as being a member of that population.
There's no bias at all in this random sampling because you
are selecting on the basis of a random probability.
It actually is the gold standard in sampling because you have
the best chance of representing the population
if you have the adequate sample size in your sample of the population.
But it's very difficult to achieve, very difficult to achieve.
Sometimes, for example, you might just say,
well I'm going to use a random selection of numbers in the telephone book,
in doing a telephone survey, but the assumption there is that everybody
in the population you're interested in has a telephone.
And it could be that people have abandoned their telephones in there homes,
because they using cellphones and they're not representative in the same way in
the particular directory you are using to get to telephones.
So even with a random sample that you assume it's random,
sometimes the selection process is such that it is not really random.
It's very difficult to achieve.
The second is a stratified sample.
Again, it's a probability sample because you can divide the population into
relevant sub groups, typically demographic categories,
such as gender and social class and education level and religion and so forth.
And then you take your sample has the same probability that you find in
the population.
So if 42 percent of your population is college educated then 42 percent of your
sample should be college educated.
So you're not doing a random sample of the population as a total random but
you are taking a sample that equals the population in
terms of the strata that you use in picking the people in your sample.
Those are probability samples, but what about non-probability samples?
They're not truly totally representative the, but they're easier to obtain.
It is easier to get to but we do worry about that error
that we have because of the being a non-probability sample.
So we try to approximate the probability sample as best as we can
by reducing this bias, reducing the error that we have in that sample.
And there are three types as I said.
There's quota sampling, purposive sampling, and convenience samples.
Now quota samples are like stratified samples in that we try to come up with
matches of probability of certain subgroups.
But it is non-probability.
We don't necessarily have all of the strata identified that are necessary to
make it a probability sample.
We might say, well, I want to make sure I have 60% women and
40% men if I'm trying to have a population of people over 65,
because there more women over 65 than there are men.
So as quote, I'm using a quota in the subgroup that I'm actually measuring.
Purposive sampling is where the population characteristics are unknown, so
I can't do stratified or quota sampling.
But I have a particular group I'm interested in.
You see that a lot in political polling.
I want to know what uneducated white men above 45 years of age are going to vote.
So your sample is a purpose of a particular group,
subgroup of the total population.
For difficult to identify groups,
let's just say I want to look at people that are homeless.
It is very difficult to say, well, here's not homeless group and
I can sample from that.
We sometimes have to use snowball purposive sampling, where we use
the people we've already measured that give us other people from the same group.
So we sort of get that sample by snowballing from the people we've
already measured for these groups that are very difficult
to identify both the population that we can get the sample from or
the sample because we just don't know how to find these people.
Most of the time, we use convenience sampling, sometimes called accidental
sampling or opportunity sampling, or volunteer sampling or haphazard sampling.
It's the sample that we have easily available to us.
For example, most psychological research done by
psychologists in academic settings use college students.
And we can't really say that college students are representative of
the population at large.
And most of the time we don't want to just use the population to be college students,
because we don't want to say that college students represent some larger group.
But that's who we have available to us and that's who we use.
Or interviews where we just go up to the man on the street interviews as
they're called.
Whoever happens to walk by as we're standing out on the street doing
interviews or the people that we use, and that sample is a convenient one to use but
it might not be representative, it's not a non-probability.
Now remember, what we're trying to do now is take a sample and
make inferences from that sample about the population at large.
Sometimes that's easy if we have random sampling.
Sometimes it's very difficult when we have to use convenient sampling.
Thank you.
[MUSIC]
