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 1E: Quasi-experimental Research
Loading...
From the course by Georgia Institute of Technology
Experimental Research Methods in Psychology
ratings
From the lesson
Introduction
Meet the Instructors
Dr. Anderson D. SmithRegents’ Professor Emeritus
School of Psychology
Hi, Anderson Smith again, and
today we're going to talk about quasi-experimental designs.
In talking about the experimental methods, we talked about how important
it is in all experiments that participants be randomly assigned to the conditions,
the different groups in the experiment.
So you're no more likely to be in one group than be in another group.
But sometimes that's impossible to do, we simply cannot do that because of
the nature of the questions we're asking in our research.
As I've mentioned I study ageing, adult ageing, it is impossible for
me to do random assignment to groups.
I can't assign someone to a group because they already have their age, and
so they are already assigned to either an 80-year-old group or a 20-year-old group.
You find quasi-experimental designs in childhood aging studies,
in other studies of adult aging like my own, in gender comparisons, I can't
randomly assign someone as a female or a male they either are or they are not.
Lots of studies dealing with brain damage, like the frontal lobes for example
are very important in making decisions and what's called executive function.
And one of the researchers has helped us understand the frontal lobes and
executive function is where we compare people that have frontal damage and
frontal lesions due to stroke, or damage, or whatever, and
people that do not have frontal lesions, and we look at how their behavior is
different, tells us a lot about how do frontal lobes work?
Or split brain studies, we have two hemispheres,
those two hemispheres are connected by the corpus callosum and
the two hemispheres have very different functions, asymmetry as it's called.
And so some people that have severed corpus callosa,
have damage in the corpus callosum, do not have the ability for the two
hemispheres to speak to each other, they're called split brain patients.
And so we learn a lot about how the brain communicates by comparing split
brain patients to people that do not have lesions in the corpus callosum.
Now, they're not randomly assigned, we cannot do that.
And we don't want to do studies which do allow us for random assignment
like, okay, you're going to be in a split brain group and you're not going to be in
a split brain group, wait around until we cut your corpus callosum.
We're not going to do that, so we have to compare groups that already exist and
that's called quasi-experimental designs.
In all of these studies, all of them,
we're not talking about random assignment, let's take an example.
If you're an 8-year-old or a 12-year-old, you can't be assigned to a group.
You're defined by your age itself, you're either in the 8-year-old group or
you're in the 12-year-old group.
And when we have those kinds of problems
we can deal with them by using different kinds of designs.
The first is called longitudinal research,
where we can simply take a group of 8-year-olds and wait four
years until they're 12-year-olds, compare the performance in the same group.
And now we have longitudinal research, we're looking at the same people and
so we avoid the whole notion of random assignment.
The same people when they're 8 years old and when they're 12 years old, and
that's easy to do with young children, it's almost impossible to do,
if not impossible to do, in old adults.
You've gotta wait around a very long time to do longitudinal research, so
we rely on cross-sectional research where we simply compare 8-year-olds to
12-year-olds, or I'll compare 25-year-olds to 75-year-olds.
And that's called quasi-experimental because now we can't assign people to
the groups, they are already assigned.
Now longitudinal research we're looking at age changes,
which say we want to do, rather than just age differences.
because age differences could be due to all kinds of things, but
age changes refer to what happens just due to aging.
These kind of longitudinal studies with adult aging are very rare because it
costs a lot of money and it takes a lot of time.
There are very good longitudinal studies of aging but they're very rare,
and they're very large, and it takes a lot of time,
you have to wait around decades of years to get the research that you want to get.
With longitudinal research you have other problems, for example we have attrition.
With aging research people die, people say I don't want to do this anymore, goodbye.
And that's okay, but that means we don't have the person in the 50-year-old group
when we had them when they were in the 25-year-old group,
we have differential attrition.
We also have practice effects, I test your memory when you're 20, I test it when
you're 30, I test it when you're 40, you simply could be learning to do better
on that task because you've done it three or four times, so that's a problem.
And then the third is just sometimes what we're measuring loses its interest
because of the time it takes to wait to do it in aging research.
An example of that would be if I'd done a study of short-term memory
in my memory research 40 years ago when I started my research career,
I would have measured digit span, just how many numbers can you remember,
now I know that digit span is not really a measure of anything that's meaningful.
A better measure of short-term memory would be what's called working memory,
where I have you learn a series of things, remember a series of things and
repeat them back, while you´re actively processing like you do when you´re
listening to me in language.
You're listening to me, you don´t know how this sentence is going to start, I don't
know how this sentence is going to end, yet I'm talking and you're understanding.
Because you go back into your memories, you find the words, you put some together,
and you have to remember the words long enough to make a sentence and
understand the language that I'm speaking.
So short-term memory, while it's short-term, is a very active process and
not just passively how many digits can you remember if
they're repeated to you one at a time.
So we have attrition effects, we have the practice effects to repeating
the test over and over again, and then we have this problem of measuring a construct
that might be meaningful in 1970, but not meaningful in 2018.
Now cross-sectional research if I'm doing aging is also quasi-experimental research.
The problems with cross-sectional research when I deal with quasi-experimental is we
have this concern that the treatment group and the control group are not comparable
at baseline, and we know they're not because they have different ages.
Also we have this problem of when we look at different age
groups that they come from very different birth cohorts.
People that I'm measuring at age 50 now
are very different people than I measured at age 50 in 1970,
they were born in different times and they had different experiences.
For example look at what's happened with prenatal care over that period of time, or
overall healthcare.
Tremendous advances in how people are treated and what diseases,
we had diseases when I was growing up that people don't even have diseases now,
those diseases have been eliminated.
We also have educational experiences, the kind of experience you had when you went
to fifth grade are different from the experiences my grandchildren are having
when they're in fifth grade, so those have changed.
And different birth cohorts have these differences already existing,
also just diet changes, lifestyle changes, all of these happen.
Remember, you didn't have electronic devices to look at
all the time when you were born in 1945, but you do now.
So there are many confounding variables when you have cross-sectional or
quasi-experimental research when dealing with adult aging for example,
lots of them.
And I want to point out there are some very good large scale,
longitudinal studies that are looking at people as they age.
The Baltimore Longitudinal Study
controlled by the National Institute on Aging out of Baltimore.
There's the Seattle Longitudinal Study that Warner Schaie has been involved with.
And there's one in Sweden that's a very major longitudinal study that looks
at people as they age through their lifespan.
And we've gained lots of very good information about
that as we have studied them over the years.
Just to give you a couple of examples of those,
this is from the Baltimore Longitudinal Study.
Where we're looking at memory, in this case it's errors in doing
a paired-associate task which is I give you a word and
you give me the word that goes with it.
And then I give you the pairs and you study them, and
then give you the test over.
And then look at errors, and you can see the errors go up between age 40 and
age 50.
The errors go up between age 55 and 65 in cross-sectional comparisons, that's
where I'm comparing people at age 40 to people in their 50s at the same time.
These as quasi-experimental cross-sectional studies, 55 and
65, and I see this increase in errors pretty systematically.
What happens when I look at longitudinal data?
With longitudinal data you get the same effect, which is interesting.
So here we have a case that birth cohort with this kind of memory task in this
kind of situation doesn't really have much of an effect.
Because I get the same effects with the longitudinal studies, and
these are the same people being measured at 40, and 50, and at 55,
and 65, not different people, and they get pretty much the same effect.
A more recent longitudinal study of memory was just published by Ronnlund and
his colleagues from Sweden in the famous Betula Swedish Study.
And he's looking at episodic memory, which is memory from things in your past, and
semantic memory, which sort of world knowledge, and he replicates what we found
with cross-sectional studies up until sort of late age.
And that is that episodic memory goes down significantly, while for
a large part of the lifespan the semantic memory stays pretty normal, you can
see it doesn't really change that much up until late age when it starts declining.
But semantic memory is maintained better than episodic memory.
So we do see changes in late age with this longitudinal study that I don't see In
cross-sectional research, but for most of the lifespan we simply replicate
what we see with the quasi-experimental developmental cross-sectional studies.
So in summary there are challenges with validity because
quasi-experimental designs lack random assignment to groups, but
with certain research variables it is very difficult to do that, like adult aging.
And what we see is we still have understood a lot about
what adult aging is all about and
memory performance by looking at these quasi-experimental studies.
And remember, it's also other kinds of phenomena being studied like brain damage,
how the brain works, gender, educational level, and so
forth, that can't provide random assignment.
But we still do those experiments, we have to understand the challenges
because we can't randomly assign participants.
But by comparing longitudinal studies to cross-sectional studies,
as we did with memory, we can better understand what those memory differences,
produced by different designs, actually are.
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.
© 2018 Coursera Inc. All rights reserved.
