Latest / Professor Peterson’s Psychology Classroom / Behind the Numbers: Hypotheses, Errors, and What Statistics Really Tell Us

Transcript

Behind the Numbers: Hypotheses, Errors, and What Statistics Really Tell Us · Professor Peterson’s Psychology Classroom

0:05
Welcome to Professor Peterson's Psychology Classroom,
0:09
connecting the science of human behavior to everyday life.
0:13
Here's Dr. Curtis Peterson.
0:17
Quantitative methods, we have basically 2 kind
0:21
of hypotheses that we usually state. Okay,
0:24
and if you remember, we divided quantitative methods
0:28
into descriptive methods and experimental methods.
0:33
Descriptive methods are when
0:39
we just want to describe data.
0:45
Oops, one second.
0:52
And/or we want to know the relationship between 2 given variables.
0:57
Okay, and we want to see how they're related. When we're doing
1:01
the experimental method, we're manipulating some
1:04
variable to see if it causes an influence
1:08
on another variable. The variable that we manipulate
1:13
is called the independent variable. The variable that we
1:16
measure to see if it's changed is the dependent variable.
1:25
Whoops, whoops, let me go back.
1:33
And we have 2 basic types of hypotheses.
1:36
One of the interesting things that we do as scientists
1:41
when we're using quantitative methods is we actually
1:45
try to prove ourselves wrong.
1:49
We try to prove what we think is going on is not correct.
1:52
So The hypothesis
1:56
that we really investigate is what's called the null hypothesis.
2:01
And in the null hypothesis, what we're saying is that there's no association
2:06
between variables. Okay, so let's give a basic
2:10
research descriptive example. If I want to study whether or not study
2:14
time influences grades, so as a person,
2:18
the amount of time someone studies goes up the higher their
2:22
grade is. My null hypothesis would
2:26
be that there is no association between the amount of time
2:30
we study and the grade on an exam or a
2:34
grade in the class. The alternative
2:38
hypothesis is the hypothesis that we think is
2:42
going on in the relationship. So in, in this example,
2:45
it would be assumed that the more we study,
2:49
the better we'll do on a test. So that would be the alternative
2:53
hypothesis in this example.
2:58
The experimental research— in
3:01
experimental research, again, we're talking about whether something
3:05
influences something else.
3:08
So if we go back to the study group example,
3:13
is let's say I set up an experiment where I randomly
3:16
assign participants to study 3
3:20
hours a week, or I have another group studying,
3:23
let's go 9 hours a week. So they're studying 3 times the
3:27
amount as the one group. Okay?
3:30
And what I'll want to do is then I'll have them study for something
3:34
over several weeks, and then I would give some type of test or exam.
3:39
And what I would want to know is, is there a statistically significant
3:43
difference between the 9 hours of studying versus the 3 hours
3:47
of studying? In the hypothesis,
3:51
the null hypothesis is that there would be no differences
3:55
between study times and its influence on test grades.
3:59
And then of course the alternative hypothesis would be is there
4:03
would be a difference between individuals who studied 3 hours
4:07
a week versus individuals who studied 9 hours a
4:10
week. So that's kind of how we set up
4:15
experimentations when we're dealing with quantitative
4:19
methods. And as I said, when we have
4:23
results, let's say that there was a significant difference between the
4:28
2 groups, the 3 hours for the 9 hours. I wouldn't say my null
4:32
hypothesis is correct— what,
4:35
my alternative hypothesis is correct, excuse me. What I
4:39
would say is I'm rejecting the null hypothesis.
4:43
Okay, because What I'm stating is
4:47
that there is differences,
4:50
but I'm not going to say that I've completely
4:55
proven that study time influences
4:58
test grades because then what would need to happen is this would need
5:02
to be retested somewhere else by another experimenter using
5:06
the same methods. So now
5:10
when we're dealing with quantitative methods, we run
5:14
the chance of having what's called errors or Type I
5:18
error versus Type II error.
5:21
Type I error is incorrectly rejecting the null hypothesis
5:26
in favor of the alternative hypothesis. So a Type I
5:29
error is when I say that there was a difference
5:34
between 2 groups, but there truly wasn't. It was
5:37
just a statistical anomaly. I had to deal with error in my
5:41
experiment.
5:46
When we have a Type II error,
5:50
this is when we accept— we
5:54
fail to reject the null hypothesis, meaning that we accept the null hypothesis
5:58
when there was actually differences. Okay, so the first
6:03
one is when I reject the null hypothesis
6:06
but there was actually differences. And the second
6:10
one is when I reject the null hypothesis when there wasn't
6:14
any actual differences between the groups.
6:18
And a large part of the reason why we care about these
6:23
errors is because— these types of errors is because
6:27
there's always noise, what I would call noise in our data.
6:32
So there's in our experiment There might
6:37
have been somebody in the 9-hour— a group of individuals in the 9-hour
6:41
study that didn't quite actually do the full 9 hours of
6:44
studying. And that would introduce error into
6:48
my experiment. We're going into— again, we're going
6:52
into biological psychology.
6:58
And one of the concerns with biological psychology is
7:02
that when we're dealing with, for example, brain imaging,
7:08
A lot of the experiments are done because it's so expensive
7:12
to do those types of experiments. Your sample size
7:16
is usually 7 to 9 people,
7:21
and that's about it.
7:24
So for a lot of experiments, when they come
7:29
out in the news and they say the
7:35
limbic system, this little part
7:39
of the limbic system controls anger,
7:42
for example. And this is just an example, okay?
7:47
Well, a lot of times you won't ever hear that again because
7:51
what they're saying is that that area represents everybody
7:54
based on a sample of about 7 to 9 people.
7:59
And usually what happens in these experiments is they repeat them
8:03
And they do them over and they don't find any results.
8:07
And it usually has to do with the sample size. So when
8:11
dealing with imaging studies, we usually use what's
8:16
called a meta-analysis technique, where instead of
8:19
looking at a single experiment to see if there's
8:24
results, we take all the experiments that are similar
8:28
to each other and we smush them together. And then that creates
8:33
a large enough sample that we can actually say something reliable
8:36
with it. Okay, so those are just some
8:40
examples of type of errors and what we
8:44
have to look out for when we're, when we're really interpreting data
8:48
that comes out. We're not really going
8:52
to spend too much time on sampling techniques, but I do want to talk about
8:57
I actually want to go on here.
9:06
Okay, what I want to start with is some concepts
9:10
in quantitative methods
9:14
that's kind of really important to understand.
9:19
Most statistical methods assume that data is
9:22
what's called normally distributed, and we'll get to this
9:26
if you've ever heard of the bell curve.
9:30
What this is, is when you sample something
9:34
long enough, it will create
9:38
what's called a bell curve. And the bell curve looks like
9:41
this, this figure right here, where if we draw a
9:45
line down this middle, that's going to represent our mean
9:49
or our average, where 50% are on one
9:52
side and 50% are on the other. But what the bell curve does
9:56
is it tells us something about the population.
10:01
So when we look at the mean, for example, of a normal distribution,
10:06
50% lays on one side, 50% lays on the other.
10:10
But then we have something called a standard deviation.
10:14
Okay, standard deviation is a measure of what we call variation,
10:19
the variation between scores. Okay, and what the
10:23
basic definition of a standard deviation deviation is
10:28
it's the average distance between individual
10:32
scores and the mean. Okay,
10:35
so it's telling us what the average differences between
10:40
individual scores are. Okay, well,
10:43
when we set the standard deviation, which is just
10:47
a representation of variation on the normal distribution,
10:51
we can tell a lot about the population.
10:55
For example, we know that
11:00
when we look at both directions from the mean,
11:04
that in the population, 68% of people
11:09
score between 1 standard deviation of the mean,
11:12
as you can see right here.
11:19
We know that 68% of people will lie between
11:23
these 2 scores. We also
11:27
know that 95% of people
11:32
lay between 2 standard deviations and
11:35
then the rest go beyond
11:39
that. So the other 4% are beyond that
11:43
in the third range. Okay, now how is this
11:47
stuff kind of helpful? Okay, well, when we look at,
11:51
for example, intelligence tests, Or IQ tests.
11:56
Most IQ tests are standardized at
12:01
a mean score of 100. I wonder
12:06
if there's— no, there's not. And the standard deviation
12:10
is 15.
12:14
So when we go this direction, we'll just go 1 standard deviation,
12:18
it would be 85, right? 100 minus
12:22
15. Yeah. Okay.
12:27
And then above that is 115.
12:32
Really bad writing. Kind of. That says 115,
12:36
I promise. And so we know that
12:40
68% of people's IQ is somewhere between
12:43
an 85 and 115.
12:48
When using IQ tests to say people
12:52
who are super
12:56
intelligent, if you want to put it that way, or people who might have developmental
13:00
issues, is we start looking beyond this.
13:05
If someone scores below an 85, we'll do further testing
13:10
to see if they have some type of developmental cognitive
13:14
issue. When people score above a 115,
13:18
then we start looking to see, well, why is it that they are scoring
13:22
so much higher than the average population?
13:25
And so this is kind of what's unique about the standard
13:30
bell curve is that it can really tell us a lot about what's
13:34
going on in the population. But as I said, it usually
13:38
takes a lot of sampling.
13:41
When we look at— and the bell curve occurs anywhere,
13:45
even in nature. So for example, when a biologist
13:48
measures the growth of a plant,
13:51
If they measure the growth of that plant enough, it will
13:55
fall along this normal curve where some plants will not grow
13:59
too high, others will grow too large, but 68%
14:04
of them will grow within a certain range.
14:07
Okay, we know this with animal growth and
14:11
the like. So this is kind of the use of it,
14:17
of that normal curve that kind of gives us a description of the population.
14:22
Real exciting stuff, huh? All right,
14:26
so that's the normal curve. So the other part of it is, is we
14:30
want to know what type of variables that we have.
14:33
So I mentioned independent variables and dependent variables,
14:38
and again, independent variables are those variables
14:41
that we manipulate, and the dependent variables
14:45
are those variables that we measure. To see if
14:49
that manipulation had an effect.
14:55
There's also something called mediating variables and moderating
14:58
variables, which we'll get to. Just keep that
15:02
in mind. And then confounding and control variables.
15:06
Confounding variables are just those variables that in an experiment
15:10
I can't control as an experimenter. So for example,
15:15
I can't control anything
15:19
that you went through before you came into the lab.
15:22
There's not everything I can control
15:26
in a laboratory or a natural setting. So those are all confounding
15:30
variables. Control variables are
15:34
those things that we want to not have an
15:39
effect on our research. So for
15:43
example, some very common control variables when
15:46
we're studying things like cognition or emotions
15:50
or something like that can be things like age.
15:56
Trying to think of the other ones. Gender, those types of things.
16:00
So we'll put in a control for those so that we can— so
16:04
we're not saying that something is due to gender or due to
16:08
age or something like that.
16:13
And then we have 3 types of scales. A nominal
16:17
scale is things like yes
16:21
or no, male or female. They're categorical
16:25
in nature. When we put things in age groups,
16:28
they're nominal. Ordinal and continuous
16:33
are ones that are put on a scale
16:37
from, for example, strongly agree to strongly disagree,
16:41
measurement of just age, height,
16:45
All of those things that can go from 0 to some upper
16:49
limit are what we consider ordinal or continuous.
16:54
These are important because we
16:58
have to analyze these slightly different. And we can't
17:03
use regular statistical methods when we have nominal
17:07
data because it just skews everything.
17:12
So there's some statistical analysis that are most commonly used
17:16
in psychology, which are descriptive methods, because a
17:21
lot of what we study we can't really experiment
17:25
on. So for example, if I want to study
17:29
the environment on development, right, I can't
17:33
ethically take a child at birth and put one
17:37
in a really good environment and one in a really
17:40
bad environment to see the success or
17:44
the development of the child. That would be unethical. What we
17:47
really rely on a lot are descriptive methods such as correlation,
17:52
regression. It's important to go
17:56
over these methods really quick.
18:00
When we're talking about correlation, what we're
18:03
talking about is whether 2 variables are associated with
18:07
each other. If we look at these diagrams in this slide,
18:14
we want to know if something is positively correlated,
18:18
which is this first one up here,
18:22
or whether something is negatively correlated.
18:27
When we're seeing if something is positively correlated,
18:30
it means when the scores in one measure go up,
18:34
the scores in the other measure go up.
18:37
So in our example, it's study time and
18:40
tests. So the amount of study time goes up, we would
18:44
expect the the test
18:50
scores to go up. I forgot what we were measuring.
18:54
A negative correlation is when the scores in one variable
18:59
go up, the scores in the other variable go down.
19:04
And then we can have spurious relationships, which is this one.
19:07
So this one's probably more curvilinear. When we're dealing with correlations,
19:12
we're only talking about linear analysis. And then this one here,
19:17
this one represents where there would be no correlation because you
19:20
couldn't fit a positive or negative line like you can
19:24
with these here. Okay.
19:28
So we want to know direction with correlation, but we also want to
19:32
know the strength of the relationship.
19:36
And how correlation is measured— and this isn't really that important—
19:42
is it's measured using that formula on this scale here.
19:45
It goes from -1 to +1.
19:50
If a correlation is 0, it means that there
19:54
is no relationship between the variables.
19:58
So if we did a scatter plot, it would
20:01
be this one right here. Okay? Meaning that there's just
20:05
no relationship between them.
20:09
A perfect positive 1 means that the
20:13
relationship is perfect, meaning as one goes up, the other one
20:17
exactly goes up in tandem. A negative
20:21
1 means that there's a perfect negative relationship.
20:26
But then the strength of the relationship can be determined between
20:29
0 to positive 1. So we know direction,
20:32
but we can also know the strength of the relationship.
20:37
And this is a good part where I would
20:41
encourage everybody to become good consumers of information. When you hear about
20:46
Coca-Cola is associated with cancer or something like that,
20:50
Okay, I heard that a lot when I was a kid.
20:54
I don't know why I always remember that, but— or you hear something
20:58
about, you know, school shootings
21:01
are associated with mental health or something like that.
21:05
Okay, is that the majority of correlations
21:10
we get in social sciences is
21:14
somewhere between a 0 and a 0.2. Or -0.2,
21:19
meaning that there are so many other factors
21:24
that go into that relationship that that
21:28
relationship usually is very weak. But in popular
21:32
media, because this would be publishable, of course,
21:35
because it'd be significant, we go
21:39
and think that those associations are one-to-one.
21:44
But in most of social sciences, we actually have very weak correlations or
21:49
very weak associations. And so when consuming
21:53
this, always kind of look at that correlation coefficient if it's reported and
21:59
keep that in mind.
22:05
The other type of associative analysis we can do is
22:08
what's called linear regression. And this is
22:12
when we want to take a predicted
22:16
variable and see if it is predicted
22:21
by another variable. So in this
22:25
example, we can go back to the
22:30
study example one. Maybe I want to run a regression analysis on
22:34
the amount of time someone studies and what
22:41
their predicted score on an exam would be. That's one thing
22:45
I can do with regression analysis. And so my
22:48
predicted variable in this case is exam scores.
22:52
That's the y-hat. Okay.
22:56
My variable that I think
22:59
predicts test scores would be study
23:03
time. Okay. So this is time and that's the—
23:06
this little b or what's called the beta.
23:11
Okay. The x is
23:16
what we call the slope of the line. Okay.
23:21
The a is what we call the constant.
23:25
So that's where it intersects our y-axis.
23:29
So it's always going to be constant. So regardless of
23:33
amount of study time, you'll always score a score of something.
23:37
And that's the y-axis. So what
23:41
I do to analyze this is let's say I
23:45
have a person who, who studies for 6 hours
23:49
a week. What I would do is I would put the 6 hours a
23:53
week here. I would multiply it by the
23:57
slope of the line, which is this, this line here,
24:00
and then I would add the constant. And what
24:05
that would give me is a predicted score that that person would
24:09
score on an exam.
24:12
This is regression analysis. And really,
24:15
because of error and everything and because there's really no perfect relationships,
24:20
what we want to know is what is the strength of the predictability
24:24
usually.
24:30
Okay, so again, the power of regression
24:34
analysis is it tells us how much one variable
24:38
can predict another variable. Correlation shows us
24:42
the strength of relationships.
24:47
So I also mentioned that there's 2 other types of
24:51
variables. One is a mediating variable and one is a moderating variable.
24:57
In mediation,
25:01
what we want to know is if there's variables between 2 associations.
25:05
So again, we'll go back to study time. So this would be
25:09
study time and this would be exam scores.
25:16
With that little E kind of. And so let's say
25:20
we found an association between those 2.
25:25
What if I think
25:29
that there's some type of something that mediates study time with exam scores?
25:35
Let's say
25:39
location where the person studies or something like that. So school or home
25:44
or in a group or not in a group. Let's go with that.
25:47
So how many people I feel people study with
25:51
mediates the relationship. Well, what should
25:55
happen is if it is truly a mediating variable,
25:59
what it should do is lower the relationship
26:03
between study time and exam when I put this in the middle
26:07
of those 2. So if we did a graph, kind of,
26:13
and let's say this is the association between study time
26:17
and exams. What I would expect to happen
26:22
is if I put in— into
26:26
my statistical software, if I put
26:30
the number of people that the individual studies with, what I
26:33
would expect is this association to go down.
26:38
And that's what mediation means, is it weakens
26:42
the relationship or it changes the relationship between
26:46
your 2 original variables.
26:51
Moderation changes the
26:55
relationship completely. And this actually comes
26:59
from a real experiment where there's moderation.
27:03
So mediation just changes the relationship. It either weakens
27:07
it or strengthens it. Moderation makes the relationship
27:11
go away completely.
27:14
Mostly. In one study,
27:18
they found that stress is
27:23
associated with depression. The amount of stress
27:27
someone was experiencing was associated with depression.
27:35
What the experimenters thought though is that
27:38
the moderating variable between stress and depression would
27:44
be something like social support. How many people does that person
27:49
have to rely on in their social system?
27:53
What they found is that, as you can
27:57
see in this graph, when there
28:01
was no social support, there was a strong relationship
28:05
between stress and depression.
28:09
But once you've had individuals who had good social support,
28:14
the relationship between stress and depression went
28:22
That has then informed us about people
28:26
who are heavily stressed and who are also experiencing
28:30
depressive symptoms. Because if we know, if we
28:34
get them to engage their social system, if we provide social support,
28:39
stress will no longer be a variable that
28:43
will be influencing their depressive symptoms.
28:46
If their depression continues after implementing some
28:50
type of social support program, that means the depression
28:54
is associated with something else besides stress.
28:59
Does that make sense? Okay with that?
29:03
Okay.
29:05
I'm not going to go through that.
29:12
Okay. Now, the other thing that I do want to talk about is in a
29:16
lot of research, when you hear the term statistically significant,
29:23
okay, everything is based upon probability
29:26
of a result. Okay? And in the field of science,
29:30
not only psychology, biology, physical science,
29:34
we always allow a little bit of error. Okay,
29:38
and so we set our probability value, which is called
29:41
the p-value, at less than 0.05,
29:46
meaning that when I run some analysis of either difference
29:50
or a correlation, I will say that that result
29:54
is statistically significant if the
29:58
value I get is less than 0.05,
30:02
which means that there's less than a 5%
30:06
chance that the results I obtained
30:11
were obtained by error,
30:15
meaning that it was some flaw in the research design.
30:23
But what does that p-value mean? It really
30:27
means the probability that those results can be replicated.
30:31
The problem is, is it's based upon the same
30:35
population. So if I I go back and resample
30:41
the population that I did my research on,
30:44
that's the probability that I'll get those results again.
30:48
Okay, the mistake a lot of people make,
30:51
and this again goes back to the importance of reading psychological
30:56
research or social science research, is that we assume,
31:01
especially when someone says this was statistically
31:05
significant, that that then applies to
31:09
everybody. And that's not what
31:12
actual p-value means. It means that if I go resample the same
31:17
people I just sampled, I'll get those results.
31:21
And so when you hear the term statistically significant,
31:26
know that that doesn't mean that it might apply to
31:29
all the population. In fact, what we should do is
31:33
go back and read the sample that it came from and
31:37
to see what are the races, what are the gender distribution,
31:42
what is the even education level and all
31:46
those, because that is what that probability represents.
31:51
And this has been an issue in psychology over the
31:56
last about 20, 30 years because we're having what's called a replication
32:00
issue. In that a lot of people are starting to
32:04
replicate really traditional research that has been done
32:09
with different populations, and we're finding that there's
32:13
differences, that the results don't replicate.
32:17
And indeed, this kind of makes it easy to study this class.
32:21
About 90% of statistical research has not
32:25
been replicated. It's not replicatable.
32:29
Because we relied on what's called this p-value.
32:33
Okay,
32:38
so what we— what a lot of good scientists have started
32:43
to do is we've added the p-value, but we've started
32:47
to look at something called effect sizes. Okay,
32:51
an effect size— so when we're looking at differences in
32:55
data, Effect size tells us how large
32:59
of a difference there is. When we're looking
33:03
at associations such as correlation, it tells
33:06
us how much one variable can
33:10
account for another. And let me kind of explain
33:14
why this is important.
33:20
So there has been— I think there was a book written years
33:24
ago called Men are from Mars and women
33:29
are from Venus. Because what they argued
33:33
is that there's such big differences between males
33:37
and females that we're
33:41
2 foreign entities almost. I'm going
33:45
to go back here and use this.
33:48
Hopefully I can draw. Let's say
33:53
we're measuring some gender difference. I don't
33:58
know, masculinity versus femininity.
34:01
Masculinity being dominance, femininity being
34:05
caring and nurturing. Let's say this
34:09
graph right here, the black line represents women.
34:14
Then if we measure men, this line,
34:18
let me see if I can draw it, the blue line,
34:23
represents men.
34:32
Okay, based on this research,
34:35
because there is this difference, this mean difference,
34:40
I would find a statistically significant
34:43
difference between men and women on masculinity
34:48
to femininity. With masculinity being on this
34:51
side and femininity being on this side.
34:57
The problem is, is in most gender
35:01
research— whoops— in most gender
35:04
research,
35:09
we don't look at the effect size, how big of a difference there
35:13
is. And when we look at things like masculinity and
35:17
femininity, this would be the actual true
35:21
story. There are a few men
35:26
that are more masculine than feminine,
35:32
and there's a few women who are more feminine
35:36
than masculine. So there are a few men
35:39
who come from Mars, there are a few women
35:43
who come from Venus, but the majority of us
35:48
right here represented in underneath here,
35:52
the rest of us just come from the planet Earth, meaning that there's
35:56
so much overlap between genders that really when
36:01
we're talking about gender differences, we're usually talking about a
36:05
few subset of the male population and a few subset
36:09
of the female population.
36:13
That looks like something my grandbaby drew this last weekend.
36:18
Anyways, so that's what effect size tells us, and it
36:22
gives us a better prediction of what actually is going on
36:26
in a population and how much true differences there are
36:31
in a population. In the correlational example,
36:35
and this is something that does come out of our example
36:39
of study time and test scores,
36:43
is if we run a a power analysis or an effect
36:48
size analysis on that, we find that the effect
36:51
size is what we call 0.2— it's around 0.20.
36:57
Okay, now what does this mean?
37:00
If we convert this into a percentage, it would be 20% by
37:04
moving the 2 decimal points. Okay, and what we
37:09
would be able to say at that point is that based
37:12
on our measures of just study time and our measures
37:16
of exams, the variability
37:19
in exam scores— what someone receives
37:24
on exam— 20% of it can be explained
37:28
by study time. Okay, now that's really valuable
37:33
because that tells us that there's 80% of
37:37
something else that explains study time, that we're only
37:40
explaining about 20% using
37:44
study time. So our exam scores, we still
37:48
haven't explained about 80% of the reason why someone does good
37:53
on an exam. So this is an example of effect
37:57
size when we use it, when we're using regression or
38:01
correlation. All right.
38:05
And so that kind of gives us the full story of what's going on.
38:11
All right. Now that we've gone through all that, did everybody understand everything I just
38:16
said? You're good with it?
38:20
How about, how about, how about online? Everybody kind of follow
38:24
that so far?
38:29
Okay. And the conclusion is— and thank you, Marissa and
38:32
Heidi. And Kelly and Joe,
38:36
the conclusion is, is whenever we're consuming some type of research,
38:41
we need to look beyond statistical significance.
38:45
We need to see if they reported things like effect size so
38:48
we know how big of a difference or similarities there are. And we
38:52
have to look at where did the research come from, who was the sample,
38:57
who were the participants. Before we should make conclusions that
39:01
something is universal or something applies to everyone.
39:04
Okay, and so those are kind of the takeaway points.
39:08
All right, all right, so that is— excuse
39:12
me, Professor? Yes? Um, with these studies that,
39:15
you know, you're obviously looking into for research, um,
39:19
would time also be implemented in this? Because,
39:23
you know, it happened say the research happened back in the
39:27
'80s or '90s, would that carry over as well,
39:31
or would we have to look for a newer
39:36
form of research in order to, I guess, come up
39:40
with our solution?
39:44
Sometimes yes and sometimes no, because we assume time
39:48
is a variable. And so yeah, in some of this research, it's,
39:51
it's whether it was replicated or not,
39:54
we, we may think it's due to time, we may think it's because
39:58
we used a different sample or something like that. But you
40:02
bring up a really good point about whether something is universal
40:06
or not, whether or not, you know, what, what happened to people 40,
40:10
50 years ago still applies today.
40:14
And there's some really good examples of this,
40:17
um, in social psychology. So Last
40:22
week we watched that video with Philip Zimbardo and he went over the Milgram
40:27
studies as an example. And every
40:30
generation that comes along in the field of psychology assumes
40:35
that my generation would not be
40:39
as compliant as previous generations, that back in the '60s
40:43
there was much more compliance
40:47
to authority, much more respect for authority.
40:50
And so psychologists have the habit
40:55
of trying to replicate that research every 10,
41:00
20 years. In fact, the last replication I think was done in 2018.
41:06
Now they use a different methodology because it
41:11
was unethical what Milgram did. And so there's some alterations to that research
41:15
to make it more ethical.
41:18
But what we have found is that compliance rates to authority,
41:23
just as many people today would shock someone or
41:27
would injure an animal or would do something just
41:33
by the orders of some authority figure that they see,
41:36
that really compliance
41:41
rates have not gone down.
41:44
Now, on the other hand, there is some famous social psychological research
41:48
where you
41:54
have a group of people anywhere from 3 to about 7,
41:58
okay? And you show the individuals
42:02
some lines in one area. Here,
42:06
I'll try and— so in
42:11
one box, for example,
42:15
there's what's called a reference line.
42:19
And then in another box,
42:23
There is like 3 different shaped lines,
42:27
okay? And they're
42:33
labeled A, B, and C. I'll just do the A because I'm
42:37
not writing very well today. A, B, and C.
42:40
And the question that they ask is, is the
42:44
reference line in box 1 is
42:49
equivalent to which line, A, B, or C? Okay,
42:53
now in this research, the only participant is the person
42:57
on the end. The rest of the individuals work
43:01
for the researcher, and sometimes they give the correct
43:05
answer and sometimes they give the incorrect answer. And our—
43:09
the question we have as researchers is, what will that
43:13
person on the end answer? What answer would they give?
43:17
So if they know that, let's say in the So we
43:21
have 5 people and the first 4 say it's A,
43:27
so it's too short of a line, but maybe it's the middle
43:31
line that is equivalent. And so they want to know, will that last person give
43:36
the correct answer, which should be B, or will they conform
43:40
to the social group, which is A?
43:44
Well, in that original research,
43:47
75% of individuals initially gave
43:51
the incorrect answer because of the group.
43:55
They went with the group instead of knowing what that line really
44:00
represents. Okay, well, in this example,
44:05
this is where we have seen changes— to get to your question
44:09
in a long way— this is where we have seen changes in
44:13
people is that in the original research,
44:17
about 75% of people would give the incorrect answer.
44:22
And today, in more modern— in, in replications of that
44:25
study, only about a third, only about 33%,
44:29
which is still a lot, but only about 33%
44:33
will give the wrong answer because of that social influence of
44:37
everyone else giving the incorrect answer. So that's an example
44:42
of how time and generations can influence outcomes.
44:45
In some areas, like conformity to authority,
44:49
we don't see time changes. When we look
44:52
at things like social influence, as the example given earlier,
44:57
we have seen a shift in the way people
45:00
respond. Does that, in a roundabout way, answer your question?
45:05
Like, yes, it does. It gives me sort of an answer.
45:10
It just kind of seems like socially we're advancing,
45:14
but in other terms,
45:18
in other ways, we aren't. Yeah. And maybe
45:22
another good example is IQ scores. Okay.
45:25
Um, I don't know if any of you in this room realize it.
45:29
I, I'm assuming everyone's in here of average intelligence.
45:33
But does everybody know that you would score higher on
45:37
an IQ test than Einstein?
45:44
You guys are smarter according to IQ tests
45:50
than Einstein. You're smarter than the
45:54
guy that got hit on the apple, Newton. In fact, you have twice as much
45:57
intelligence statistically as Isaac Newton.
46:02
Because what we've seen again is a generational shift.
46:06
We see that about every 10 years IQ
46:10
scores within a generation increase between 2 to
46:14
5 points. And so Einstein scored
46:18
around 130 during his time.
46:23
While his 130 compared
46:27
to IQs— what people score on IQ scores today— his 130, he would have
46:31
scored around down to 75 to 85,
46:36
where again, I assume
46:39
because this is a bright group, everyone here scores average or more.
46:44
I'm assuming a lot more. Okay, so really when
46:47
we look at things like IQ, we see our IQs
46:51
actually every 10 to 20 years are increasing
46:55
and we have to reset the average. So that's
46:59
another example that people often don't
47:03
think about, that yes, you're more intelligent than Einstein
47:07
and Newton. You should go home and tell your parents that.
47:16
So, and Tommy, your question is very good. We really
47:21
don't know. There's been a lot of research that
47:24
suggests that, you know, maybe it's the education, it's access
47:28
to information. If you think about, you know, Einstein, he didn't didn't
47:32
have access to Google or Google Scholar
47:36
or research or the mass amount of information
47:40
that we have today. And so there's some theories and some
47:44
thoughts, but we actually don't know why IQs
47:47
have been increased like that. It's just an effect that we see.
47:51
So, but that's a good question, maybe a doctoral
47:55
dissertation in your future, maybe. So Yeah,
48:01
and I kind of think it has to do with technology.
48:05
So, okay. All right, any other questions on
48:09
the research part?
48:17
All right,
48:25
so what I would like everybody to do for for
48:33
the rest of the class, last 15 minutes, or make up the time later,
48:39
is go to— so
48:43
everybody knows, I did— I put this in the
48:47
announcements this morning. I did redesign our
48:53
Canvas page. So when you go into it,
48:58
and collapse the Getting Started, we now just have the 3
49:03
modules. And I'm hoping it'll make it a little bit easier to navigate.
49:07
So the first module is our recorded lectures.
49:10
The second module has all of our assignments that are due in
49:14
the assignment area. And then this last one
49:18
is just all the course learning materials. So if you want to go study more
49:23
about sensations and perceptions when we get to that, you can go to that.
49:27
But it also has the learning materials And what I'd
49:30
like everybody to do between now and Wednesday is go into
49:34
the course learning material and download
49:40
or read this. It's in a PDF. It's the Lieberman,
49:43
Einsberg, Williams article from 2003.
49:48
And it talks about an fMRI study dealing with—
49:52
I can't talk— social exclusion.
49:56
Okay? And kind of take a look at that because we want
49:59
to look at the measurements that they used and whatnot on Wednesday when
50:03
we get into understanding how we measure things
50:07
when it comes to, for example, social exclusion
50:11
or any of the other variables that we measure. But also
50:15
we'll have a discussion about the relation between these areas of the brain and
50:20
functioning.