0:11
Welcome back, that was a great experiment wasn't it?
So what Ash did, he really varied the number of confederates he used.
The people who were aware of his hypothesis
about group behavior and the size of groups and conformity.
He varied the number of confederates from as little as
one in the group Tasmania as 15.
So in the initial experiment the participants, the real subject and
the confederates will all sit in a room they were asked questions about
these lines, as you saw, and then they responded of course
groups were told to respond out loud.
Now, as the size of the groups increased,
Ash found more and more conformity.
If it's just with one person then the person feels no need to conform.
Two, not so much as well but the larger the group, three, four.
Well then, yes, you feel the desire to conform.
Something must be wrong with you, as the young gentleman showed in the experiment.
And not them, so you want to conform and give the correct response.
1:26
Now in an ideal experiment, you have to control
all of the variables except the one that is manipulated, all right?
But in reality, you can really directly control very few variables.
So in practice extraneous variables are not merely possible causes,
there are possible cause.
In this case it is possible to believe that the students in ash experiment
were influenced by the gender of the participants.
Or let's say, if a young person went to high school and
were to do this experiment, with all the other participants being college students,
this will also lead to them agreeing with others.
So both, say if you're a man and you're with three women, in the time that when
this experiment was done men being the sexist
folks they were back then might feel where they did not have to agree with the women.
Or, so if you have a really young student, a freshmen with seniors or
senior with some freshman students, that might influence a response as well.
So to control for that, actually use college age students,
none who were really much younger than the others.
And he tended to use individuals from the same gender, all right?
A possible but not plausible variable could be income levels.
So a poor students who may not be as well dressed could
possibly be influenced by the other students who seem to be more wealthy.
But this doesn't seem likely in this case since income may not be as
apparent as say age.
You can look at someone and tell their much older than you are.
They deserve your respect even when they seem to dispute them too much.
So, we have to distinguish extraneous variables, the age of the student,
the gender of the student, from what I call noise variables.
Variables that are not really plausible explanations.
Things like the income level of the student.
Are you with me?
Good.
So you have to look at these variables and control for
these extraneous variables that are plausible.
3:51
Okay, the next step is sampling.
So initially we talked about the research approaches, so
we talked about observations to be an experiment, alright Now we have
to discuss sampling, sampling.
So you go to the sampling unit, size, and procedure.
So the sample is really
a smaller percentage of
the entire population.
Let's think of this, let's step back.
What's a population?
The research population all the possible individuals whom
you could have serve that relate to your variable of interest.
Let's say you wanted to collect data on college students.
How do college students respond to
a reduction in the drinking age?
5:01
Let's say you want to reduce the drinking age from 21, the legal age from
21 to 19 and you want to study how students might possibly respond to that.
So, you have to survey them.
There's absolutely no way you can get every college students in the US,
no way, no way impossible.
So what do you do?
You select a segment, a representative sample, to survey.
And then you are able to predict, based on that,
you are able to generalize to the entire population.
And the more representative
the more like the general population the sample is, the better.
The better you're able to generalize.
So, you must choose a segment for cost reasons, for time reasons, for
efficiency reasons, you have to select a sample, a smaller sample.
You can't do the entire population.
All right, this sample when you talk about selecting a sample class 3 decisions.
Who is to be surveyed?
6:06
Right, that's your sample unit.
The next decision, how many people should you survey?
That's a sample size.
And finally, how should the people in the sample be chosen?
And that's your sampling procedure.
So you have to look at the sampling units, the sample size, sampling procedure.
6:28
So you have to determine who's to be surveyed
based on the information that you need.
In the example I gave, we are looking at how do we go about determining if
reducing the drinking age from 21 to 19.
How would that impact consumption of alcohol?
On college campuses with college students, all right,
in that case I need to sample college students.
These are the ones.
6:59
How many people should be surveyed?
Large samples give very reliable results but are very costly, all right?
A sample can still be small and provide reliable results, you will hear from
faculty and all these academics who do research that
the sample must be reliable.
So in order to make sure that you can do analysis with your sample.
They usually say that the sample must be 30 or more.
That's kind of rule of thumb, I don't always buy it.
I actually like larger samples, but in some of my research, for
example I'm doing functional magnetic resonance imaging research, and for
my research that's very costly, but it's also a little more accurate.
Because you're measuring blood flow in the brain, and usually you can do research
like that and get reliable results with five or ten subjects, okay?
It's difficult for people to lie when you can see the blood flow in the brain saying
they're either angry, happy, those parts of the brain.
It's more accurate research than asking questions using surveys,
so smaller samples are sometimes acceptable.
But large samples usually give more reliable results.
Sampling procedure, how should it be chosen?
There are two different ways you can choose a sample either probability sample.
This is when any member of the sample could be chosen using this method.
They have equal likelihood of being chosen, or,
nonprobability sample, which means that you choose them based on convenience.
Let's say I walk into my classroom and
I say, I have to collect data for one of my research projects.
Listen you guys all here, I'm going to use you as subjects.
9:00
So it's a non-probability sample.
I'm only choosing them because it's convenient because they are there.
Probability samples, if for this I'm sampling the population in my
research project, call it, students were probably looking at the drinking age,
to get a probability sample I should go somewhere
that would ensure that students in my,
in the population will have equal or close equal probability of being sampled.
So what could I do?
I could ask the registrar's office to give me a list of all students and
use a computer program to randomly generate names that I should contact and
send them out a survey by email.
I could send those research assistants out by the quad,
on Green Street, all over campus, equal distance from each other.
Had asked them to interview different individuals.
There are different ways to enhance the probability that you will get
any one from your population through your sampling methods.
10:09
Terms to note, one, interviewer bias.
How do you, as an interviewer impact the data collected?
Studies have shown that things like simply speaking in a deeper voice,
if you're a man, wearing a lab coat if you're a man or a woman.
Things that you do can influence responses.
People try to look smart in front of people they believe are smart.
So you're wearing a lab coat, they will respond appropriately.
10:39
Also, representativeness.
Does the sample truly represent the whole population?
Does the sample truly represent the whole population?
Now that's an old example there but let me give you an example, a current example.
Let's say I survey people in New York for the 2016 election.
And then I used that as a basis for saying, Hillary Clinton is going to win.
11:05
She's going to win, these are going away.
No I can't, because New York is definitely not representative
of the entire population of the US.
I've lived in New York, I've lived in other states, California Illinois, etc.
And New York is far from representative of the entire United States.
So we have to look at representativenesss.
Is that sample you're using, how much is it like or
how much does it reflect the attributes of the entire population?