This is a five-section course as part of a two-course sequence in Research Methods in Psychology. This course deals with experimental methods whereas the other course dealt with descriptive methods.
Lesson 1D: Control
Loading...
From the course by Georgia Institute of Technology
Experimental Research Methods in Psychology
ratings
Explore Related Videos
At Coursera, you will find the best lectures in the world. Here are some of our personalized recommendations for you
From the lesson
Introduction
Meet the Instructors
Dr. Anderson D. SmithRegents’ Professor Emeritus
School of Psychology
Hello, Anderson Smith again.
And today we're going to talk about
the most important consideration when using the experimental method in psychology,
and that is the idea of control.
Now, we need control because often when we're doing an experiment,
we don't have the exact relationship between
the independent variable that we
manipulate and the dependent variable that we're measuring,
as an extraneous variable that's really correlated with both.
And when that's called a confounding Variable.
We actually talked about confounding variables
earlier when we talked about variables in general.
But we need to examine the independent variable when
this extraneous variable is controlled for- this confounded variable is controlled for.
So let's take an image of what that might look like.
We have an independent variable that we're manipulating.
We assume that it's the variable that's causing
the changes we measure in the dependent variable.
But it might be that a confounding variable,
some extraneous variable we don't really know about is
correlated with both the independent variable and the dependent variable.
So we need to control that,
we need to control that confounding variable.
So we really are looking at the independent variable that we want to show,
is influencing the dependent variable. Let's take an example.
Polio was a disease when I was growing up that was very prevalent.
I had two very close friends that had polio as children,
and I went to school with them.
And they rode in wheelchairs or they had braces,
or they walked with crutches.
In 1949, a Doctor Sandler noticed that there was
a correlation between the incidence of polio and ice cream consumption in children,
when he assumed that ice cream consumption was really sugar consumption
and that was really what led to the increase in the risk for polio.
And in fact the public health officials in those days
actually issued warnings about sugar consumption in children,
saying that they had to watch the sugar consumption because it could lead to polio.
But it was warm weather that increased the risk of polio,
not ice cream consumption.
In fact, polio is a virus and we know that viruses are much more active in the summer.
And the summer is correlated with warm weather,
is correlated with both ice cream consumption,
and the activity of the virus,
that's the increased risk of polio.
So we had this relationship that Dr. Sandler thought
about sugar consumption leads to increased risk of polio,
but we also have a variable that wasn't considered that's correlated with both.
Warm weather is correlated with sugar consumption and
warm weather is also correlated with the risk of polio,
it's a confounding variable.
And we need to worry about controlling that so we
can really look at the relationship we're interested in,
and that now we know is the warm weather in the increased risk of polio.
But we have to control for one of the variables that
might be confounding to that important relationship.
So if we have a confounding variable,
that simply means we have lack of
control and we had to figure out how to control for that.
To control for confounding variables there are several things you can do.
Better, you can redefine the measure of your independent variable so it's not confounded.
For example, we want to change the variable from sugar consumption to warm weather,
we are better defining what our independent variable is.
Second thing we can do,
is include the confounding variable in the design as another variable.
So we actually are measuring both warm weather
and sugar consumption in the same experiment.
That controls for what is the remaining variable- what is a confounding variable.
And another way then we can do is match.
We simply can match,
select subjects so that one variable is not really of a concern anymore.
And then there are statistical techniques that can be used
called analysis of covariance that allows us to exclude certain variables.
All of these are possible methods in the experimental method that we can
use to control for confounding variables.
So the first is better definitions.
We can better specify the measure of the independent variable so
it's not confounded with this confounded extraneous variable.
So the independent variable should be,
sugar consumption and not eating ice cream.
That's the first thing, is that ice cream per se but
sugar consumption and that even Dr. Sandler realized that.
And sugar consumption throughout the year should be the independent variable.
That's the variable we're looking at not just sugar consumption,
ice cream consumption in the summertime.
So if sugar consumption leads to a risk of polio,
we really want to test that relationship.
Then we should look at sugar consumption throughout the year.
The second method of controlling is including it as a variable,
actually making it a part of the experiment.
So if we now have these two variables that are correlated,
we don't know which one is the actual cause,
then we should be looking at both at the same time.
For example, if we had
sugar consumption in warm or cool weather then we could have all four groups.
We could have a group that has high sugar consumption in the summer when it's warm,
and high sugar consumption in the winter when it's cool.
We can have low sugar consumption in the summer when it's
warm and low sugar consumption in the winter when it's cool.
Now, we can look at the effect of whether it's warm weather or not,
and whether sugar consumption is high or low.
They both are variables and I can look at them
both and actually look at interactions that we talked about.
The interaction between the two variables in these
and making it a multi variable experiment.
The third method for controlling confounding variables is matching.
We simply match the selection of subjects and
their sample that have the same one level of one of the two variables.
So we might equate our sugar consumption and only use summer when it's warm,
or compare, equate subjects also in summer and in winter.
So we have matched on sugar consumption and not matched on sugar consumption.
And then we look at the risk of polio,
and what we'll find is that when they're matched on
sugar consumption or not matched in sugar consumption,
it doesn't really matter if we are not looking at warm and cool weather.
We're matching them when we select the subjects,
we take subjects that have exactly the same level of sugar consumption,
and thus, we're not looking at sugar consumption as a variable.
And then the last method that we can use to
control for confounded variable is actually make a statistical control,
actually looking at what's called a covariance design.
And in that case,
we are actually statistically controlling for
one variable when we are looking at the other,
when there's a correlation between the two.
And that's called, Analysis of Covariance or
ANCOVA which actually allows us to look at one variable in two or more groups,
take into account the variability that the other variable has,
that's called a covariate.
So we have a variable we're interested in and then
we look at the covariate the other variable,
and we take that out when we are comparing our independent variable.
So the risk of polio is a dependent variable and sugar consumption might be
the covariate that we want to take out to see
whether sugar consumption is really the cause.
Let me use Venn diagrams will show how this effect works.
Here we have the three variables, the independent variable,
sugar consumption, the dependent variable which might be risk of polio.
And then we have a confounding variable,
the weather, and they all three are related to each other.
There is a relationship as you can see
the overlap between the independent and the dependent variable,
but there's also an overlap with
both the independent variable and the dependent variable of
the confounding variable, the covariate.
So we take that out.
We take the covariate out statistically,
it leaves us only with a relationship between the independent variable,
the dependent variable is left over.
And as we can see there's a significant level of that left over.
So if we said that warm weather was what we took out,
or we'll say sugar consumption was what we took out,
then we still have a relationship
between the independent variable and dependent variable.
So it's not sugar consumption,
it has to be something else,
and now we know it's the weather.
So there are different ways you can control for confounding variables,
different methods that we can could use,
but we need to do that when we find there is a confounding variable
that's correlated with both the independent variable and a dependent variable. Thank you.
Explore our Catalog
Join for free and get personalized recommendations, updates and offers.
Coursera provides universal access to the world’s best education,
partnering with top universities and organizations to offer courses online.
© 2019 Coursera Inc. All rights reserved.
