Week 2
mloi01
Research Design: Observational and Correlational Studies
Video Title: Research Design: Observational and Correlational Studies Originally Published: 2011 Publishing Company: SAGE Publications, Inc
City: Thousand Oaks, USA ISBN: 9781483397108
DOI: https://dx.doi.org/10.4135/9781483397108
(c) SAGE Publications, Inc., 2011 This PDF has been generated from SAGE Research Methods.
NARRATOR: Research Design-- Observational and Correlational Studies. Since the moment you
were born, you've been exploring the world around you. In a sense, you've been conducting research.
You've noticed the ways people interact with each other, the relative sizes of objects,
NARRATOR [continued]: and how the colors of nature change with the seasons. Each of us is an
amateur researcher, observing, analyzing, and drawing conclusions about everything we see. In order
to conduct a more formal study whose conclusions you can share with others, you need to apply
scientific methods to your research.
NARRATOR [continued]: Knowing about scientific research methods will also help you understand,
interpret, and be more analytical in your thinking about studies you read about in textbooks, journals,
newspapers, or online. To make sure your research is as strong as possible, let's talk about designing
your study and interpreting your results.
NARRATOR [continued]: Specifically, we'll focus on some overarching types of research studies,
when to use an observational design, along with some advantages and disadvantages, two different
types of observational design, those that you conduct in the field and those that you conduct in a
laboratory,
NARRATOR [continued]: analyzing data from an observational study, including some statistical
methods, when to use a correlational design, along with some advantages and disadvantages, how
to design and implement one, and analyzing data from a correlational study.
NARRATOR [continued]: Before we begin to explore research designs, it is important to understand
the terms "variable" and "construct." These terms are used interchangeably and are found throughout
scientific literature.
NICOLE CAIN: A "construct," which can also be called a "variable," is a topic of interest that varies
from person to person. Some examples of constructs that researchers are often interested in would
include things like quality of life, IQ or intelligence, or anxiety.
EVELYN BEHAR: Another variable that a lot of people are interested in is marital quality. [Evelyn
Behar, PhD, Assistant Professor of Psychology, University of Illinois at Chicago] Obviously, some
people will have very, very high-quality marriages, some people will have extremely low-quality
marriages, and then there will be lots of people in between those two extremes. So again, a variable
is exactly what it sounds like. It's something that varies across different people
EVELYN BEHAR [continued]: and we also call it a construct.
NARRATOR: Types of studies--
EVELYN BEHAR: In general, there are three basic types of designs in research. The first type is what
we call the "observational design." This is when we just want to know what is the basic nature of
a particular construct or a particular variable. So we might ask, what is the basic nature of marital
quality?
EVELYN BEHAR [continued]: The second type of design is the "correlational design." And essentially,
what we're asking here is, how do two variables or two constructs relate to one another? So you might
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 2 of 12 Research Design: Observational and Correlational Studies
be asking, what's the relationship between marital quality and parenting behaviors? The third type of
design is what's called an "experimental design and this is a little bit
EVELYN BEHAR [continued]: different. This really speaks to cause and effect and this is where you
manipulate one variable and you see the effect that it has on the other variable. So even though we
would never actually run this type of study, you might randomly assign people to have either very,
very good marriages or very, very bad marriages and then see what effect it has on their parenting
behaviors
EVELYN BEHAR [continued]: and on their parenting ability. We would never actually run that study
but that's what an experiment would look like.
NARRATOR: Observational Studies. Let's focus first on observational studies. In this type of study,
we're learning new information about one variable by watching, listening, and in some way measuring
or recording data.
NARRATOR [continued]: This is especially useful when it would be immoral or impossible to cause
a phenomenon to occur or when we're interested in getting preliminary information on a brand new
topic before investing heavily in other more expensive and time-consuming types of research.
EVELYN BEHAR: So let's say that you're a medical doctor and someone comes into your office, into
your practice, and the person has all of a sudden grown neon-green hair. And you have never heard
of this phenomenon before. And the next day, another person comes in with neon-green hair and you
think to yourself, there must be something here.
EVELYN BEHAR [continued]: Now, you might be tempted to run a clinical trial to try to treat the green
hair phenomenon. You might be tempted to go and do all this expensive research. But you've only
seen two cases of it so first, maybe you should run an observational study. So you might want to call
10,000 households in the United States. And when the person answers the phone, you're going to
ask,
EVELYN BEHAR [continued]: has anyone in your home developed neon-green hair? Let's say that
you find no other cases of neon-green hair. Well, now you've just saved yourself a lot of time and
money. You don't have to go and run a clinical trial. You're not going to go and do all this expensive
research. But let's say that out of the 10,000 households you called,
EVELYN BEHAR [continued]: lo and behold, there are 100 households with people who have neon-
green hair. So you can ask lots of questions but keep in mind that they're still observational. Even
if you find out that the age of onset was age 30 for every one of these cases who developed neon-
green hair, you didn't run an experiment. So you have to be very careful in drawing your conclusions.
EVELYN BEHAR [continued]: You cannot go on to say, turning 30 causes individuals to develop neon-
green hair.
NARRATOR: So we've seen that one of the advantages of the observational design is that it requires
less of an investment of money and other resources than correlational and experimental studies.
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 3 of 12 Research Design: Observational and Correlational Studies
Also, it gives us a great deal of information about a construct. This is especially valuable when there
is a brand new phenomenon
NARRATOR [continued]: that has just come into existence or that at least has never been studied in
the past. The observational design also allows us to generate hypotheses about the construct. In Dr.
Behar's green hair example, there seems to be some connection between turning 30 and developing
this condition.
NARRATOR [continued]: We can formulate a hypothesis about the connection and then test it using
correlational or experimental methods. On the downside, when we use an observational design, we
cannot draw any conclusions about the relationship between one construct and another. We also do
not find out about causes or effects.
NARRATOR [continued]: Once we choose to conduct an observational study, we need to decide
whether it will take place in the field or in a laboratory.
EVELYN BEHAR: Running an observational study in the field makes the most sense when you have
what's called a "naturalistic question," basically a question about a construct as it exists in the natural
environment.
NARRATOR: No matter what kind of study we're conducting, it is extremely important to develop a
systematic method for recording our data. For an observational study in the field, this can be achieved
by creating a check sheet with certain potential observations. Each time we see a certain behavior or
characteristic, we place a checkmark in the appropriate column.
NARRATOR [continued]: It is a good idea to develop a check sheet that lists as many types of
observations as may possibly interest us when we analyze our data.
EVELYN BEHAR: If you want to look at basic friendliness levels in society, you might literally go and
ride an elevator all day long. You could ride many different elevators in many different buildings.
NARRATOR: In this case, we probably will want to prepare ourselves for many levels of friendliness in
order to give ourselves a chance to see varying degrees of friendliness. Our check sheet may include
"Total Avoidance," "Eye Contact," "Nod," Greeting," and "Compliment." We also may want to reserve
our final column for "Other,"
NARRATOR [continued]: in case we observe something we didn't expect. Not everything can be
observed and recorded in the field, however. If we need a more controlled environment or more
sophisticated measuring apparatus, we may need to bring our participants into the laboratory.
EVELYN BEHAR: If you were interested in looking at IQ levels, obviously, you can't just go out into
the field, look at someone, and assess what their IQ is. You need to bring them into the lab. Have
them undergo an entire procedure where they take an IQ test because in order to do this type of
research, you have to have a controlled environment.
NARRATOR: When we conduct observational studies in the laboratory, it is somewhat easier to
record our measurements on a computer than it is when we are in the field, though sometimes, we
may be more comfortable working on paper and later transferring the data into a program that can
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 4 of 12 Research Design: Observational and Correlational Studies
help us analyze it.
NICOLE CAIN: In an observational design that takes place out in the field, your participants are
anonymous. [Nicole Cain, PhD, Assistant Professor of Psychology, Long Island University Brooklyn
Campus] Oftentimes, they don't even know that you're observing their behavior. So it's not necessary
to get their permission because they're anonymous. This means, though, that you can't have any
photographs of them or any video recordings and you cannot have any identifying information about
them at all.
NICOLE CAIN [continued]: In contrast, in an observational design that takes place in a laboratory,
you do have identifying information about them. So in this case, you do need to get their permission
in order to record and track their behavior. This often takes the form of what's called an "informed
consent form for research." This is an explanation of the study,
NICOLE CAIN [continued]: along with an explanation of the risks and benefits to participating in the
study and an explanation of how you plan to keep their data confidential.
NARRATOR: The more careful we are about systematically collecting our data, the easier things will
be when we are ready to do our analysis. Data analysis should help us describe our findings in ways
that improve our intuitive understanding of the construct. For observational studies, we usually focus
on five categories of analysis,
NARRATOR [continued]: measures of central tendency, measures of variability, kurtosis, skewness,
and shape of the distribution. Measures of central tendency and measures of variability are actual
values or numbers. Kurtosis, skewness, and shape of the distribution
NARRATOR [continued]: are often more visual in nature. Let's look at each of these types of analysis.
NICOLE CAIN: The measure of central tendency is what a typical person looks like on a particular
construct.
NARRATOR: The three most common measures of central tendency are mean, median, and mode.
The mean is the average of all scores in a distribution. This is the most commonly used measure
of central tendency. We're generally referring to the mean whenever we talk about "average" IQ or
"average" income.
NARRATOR [continued]: Occasionally, a problem arises from using the mean as the measure of
central tendency.
NICOLE CAIN: An outlier is an extreme score and an outlier can drive your mean up or down
artificially. So for example, if you were interested in looking at salaries of 300 people, you would ask
300 participants to record their salaries to get an average. But if you had one person who had a salary
of $100 million,
NICOLE CAIN [continued]: you could see how that would artificially drive your mean up. So it would
no longer be a good measure of central tendency.
EVELYN BEHAR: So the problem of the outlier is that it is an extreme score in either direction, either
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 5 of 12 Research Design: Observational and Correlational Studies
too high or too low, that essentially is changing the entire look of your sample. It's changing everything
that your sample is trying to reflect about itself. When this happens, you have two options. You can
either simply get rid of that extreme observation
EVELYN BEHAR [continued]: and that's OK to do. There's a good reason to do it. Your other option is
to simply not rely on the average or the mean as your measure of central tendency. You might choose
to rely on a different measure of central tendency, something like the median, which is defined as the
middle value in a distribution of values. So you can see that if you just line up everybody's income
EVELYN BEHAR [continued]: from lowest to highest, you're still going to have that person who's
making $100 million all the way at the right. But it's only one observation and if you're just looking for
the middle observation, that one person is not going to affect your search for the middle observation.
NARRATOR: Another measure of central tendency is mode. Mode is the most common score in a
distribution. For example, if we're observing the number of jelly beans a child eats when he has a
bowl full of jelly beans in front of him, we may find that the numbers range from zero for a child who
doesn't like jelly beans to 200 for a child
NARRATOR [continued]: who will keep eating beyond the point of getting a stomach ache but that
the most common number of jelly beans a child will eat is 20. The number that comes up the most
times is the mode.
EVELYN BEHAR: The other measure that we're interested in is variability. So whereas measures of
central tendency tell you about the typicalness of a particular observation, measures of variability tell
you about how variable your sample is around that typicalness. We may know that the average IQ of
our sample is 100
EVELYN BEHAR [continued]: but now we want to know how variable is our sample around that 100.
We might have some people at 110, 120, 130. And also on the other side, we're going to have some
people with an IQ of 90, of 80, of 70. So we're going to have people falling on either end of that
average score of 100.
NARRATOR: Variability can be high or low for a given construct. High variability means that we are
seeing scores that are way above and way below the mean. Low variability means that all of the
scores are grouped closely around the mean. They don't vary much. The most common measure of
variability
NARRATOR [continued]: is called "standard deviation." Standard deviation is the average distance
from a score to the mean, in other words, the average amount that scores in the sample deviate
from the mean of that sample. To calculate standard deviation, we first calculate the mean. Then, we
subtract the mean from each of the individual scores
NARRATOR [continued]: to find what we call "difference scores." Next, we square each difference
score. Then, we add up all of the squared difference scores. Then, we divide this number by the total
number of observations in our sample, called the "n." Finally, we take the square root of the whole
thing
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 6 of 12 Research Design: Observational and Correlational Studies
NARRATOR [continued]: and that is our standard deviation. Another thing we look at when analyzing
our data is called "kurtosis." Kurtosis is the shape of the distribution. It refers to how peaked or flat
the distribution is. A platykurtic shape is short and fat.
NARRATOR [continued]: It indicates a great deal of variability. The scores are very spread out. For
example, we may have a class of 100 students with very different levels of ability. If we administer a
calculus exam with 10 questions, we may find the following distribution. This would be a platykurtic
distribution.
NARRATOR [continued]: As our intuition will tell us, there is no real tendency here. A leptokurtic
shape is tall and skinny. It indicates very little variability. There are many scores close to the mean.
For example, we may be measuring the number of hours per day newborns sleep in the first week
NARRATOR [continued]: of their lives. Most of the numbers could be around 17 hours, with many
babies sleeping 16 and others sleeping 18 but few sleeping much less or much more than that. A
mesokurtic shape is in between. It indicates a normal or medium amount of variability.
NARRATOR [continued]: This is also similar to a normal curve, which we will talk more about later. IQ
is a good example of a mesokurtic distribution. Most people will score around 100 and scores become
progressively less frequent as we move away from the mean in either direction.
NARRATOR [continued]: We may also want to look at our sample's skewness. Skewness refers to
whether scores are distributed fairly evenly around the mean or whether there are many more scores
to the right or many more scores to the left of the mean. There are three types of distributions in terms
of skewness. A symmetrical distribution, known
NARRATOR [continued]: as a "normal distribution," is when an equal number of cases fall to the left
and to the right of the mean. The most common example of a normal distribution is IQ. The mean is
100 and the standard deviation is 15. This means that the average IQ is 100
NARRATOR [continued]: and 34% of the population has an IQ within 15 points above the mean,
between 101 and 115, while another 34% has an IQ that is within 15 points below the mean, between
85 and 99. Then, if we take it out to another standard deviation,
NARRATOR [continued]: we have about 13% of the population having an IQ from 116 to 130 and
about 13% of the population having an IQ from 70 to 84. Finally, if we go even farther out into the tails
or the almost unpopulated extremes of our sample, we will find that there
NARRATOR [continued]: is a very small number of people with an IQ that is more than two standard
deviations above or below the mean. A left-skewed distribution is also called "negatively skewed." It
means that some very low scores exist in the sample and that pushes the tail to the left, making the
mean lower.
NARRATOR [continued]: On the other hand, a right-skewed distribution, which is also called
"positively skewed," means that there are some very high scores that pull the tail to the right and
make the mean higher.
EVELYN BEHAR: One really important step in analyzing your data is to graph your data. It's an
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 7 of 12 Research Design: Observational and Correlational Studies
opportunity for you to see what your distribution looks like. And that's where you might see an outlier
in your data and you otherwise may not have been aware of it. You might see an interesting shape in
your distribution. So before analyzing anything, you
EVELYN BEHAR [continued]: should always graph your data and just take a visual look to see if
there's anything interesting or odd or wrong that pops out at you.
NARRATOR: One final way of analyzing data and learning about the nature of our construct is looking
at the shapes of our distributions. There are three basic shapes, rectangle, bimodal, and normal. The
rectangular shape comes about when all of the scores occur with roughly the same frequency.
NARRATOR [continued]: For example, if we were eating M&Ms from a bag and we kept track of how
often we picked each color, we might end up with a rectangular shape of distribution. Let's say our
bag contained 120 M&Ms, 20 red, 20 blue, 20 orange, 20 green, 20 yellow, and 20 brown.
NARRATOR [continued]: If we randomly pick one M&M 60 times, we're likely to pick approximately
10 of each color and we would end up with a pretty flat distribution, which will look like a rectangle. A
second shape of distribution is bimodal. This occurs when two groups seem to emerge from the data.
NARRATOR [continued]: The bimodal shape often can inform us about the nature of the topic we're
studying and can give us hints about sub-samples that might exist.
NICOLE CAIN: So one example of a bimodal distribution would be height. If we compare males to
females, males are always going to be a little bit on average taller than females. So you would have
one peak at the higher end of the spectrum for males and another peak on the lower end of the
spectrum for females.
NARRATOR: On a normal curve, many of the values we've measured are clustered around the mean,
with fewer and fewer cases appearing as the values spread to the right and to the left, away from the
mean. On a graph, the edges of the curve are so low that they look like tails.
EVELYN BEHAR: The age at which infants start to crawl or walk or speak exists on a normal
distribution. Intelligence exists on a normal distribution. There are many phenomena in the natural
environment and in the everyday world that exist on a normal distribution or along a normal curve.
NARRATOR: Sometimes, the information we gather about a construct in an observational study
leads us to create hypotheses that we wish to test further through correlational studies. Correlational
Studies.
NARRATOR [continued]: When we know something about our construct and we want to check on its
interplay with other constructs, we'll want to use a correlational design.
NICOLE CAIN: One instance where you would want to use a correlational design is when you want
to know what other variables are related to your construct of interest and what might be the potential
cause and effect of your variable.
EVELYN BEHAR: So let's say, for example, that you are interested in insomnia and you want to know
what causes individuals to have a lot of trouble falling asleep at night. You might first look to see what
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 8 of 12 Research Design: Observational and Correlational Studies
it co-occurs with. What are some problems or some environmental situations that tend to go along
with insomnia?
EVELYN BEHAR [continued]: And if you see that as caffeine ingestion goes up, insomnia levels go
up, then you might start to think a little bit along the lines of causation. The correlational study can't
tell you whether two constructs are causally related but it can give you an opportunity to draw some
hypotheses about causal relationships.
NICOLE CAIN: Another time that you would use a correlational design is when an experiment is not
possible. So for example, you would do this in a lot of clinical research.
EVELYN BEHAR: Let's say that you want to study the effects of depression on marital quality.
This is a great question but obviously, it's unethical to go out into the world and make some people
depressed and other people not depressed. And so in this situation, it makes more sense to run a
correlational study and simply see what happens to marital quality levels
EVELYN BEHAR [continued]: as depression levels go up or down as they naturally occur in the world.
But you are not going to run an experiment and actually go and make people depressed in order to
see the effect that it has on their marriages. You wouldn't want to do that.
NARRATOR: As with observational studies, correlational studies have some distinct advantages
and disadvantages that we need to consider when choosing our research design. As we've seen,
correlational studies enable us to begin formulating hypotheses which we can then test by means of
an experiment. They're also good to use when we can't control a variable,
NARRATOR [continued]: in other words, when we can't set up an experiment. Correlational designs
do have their disadvantages, however. It is impossible to determine based on a correlational study
whether one variable is causing the other to rise or fall. Even if we know there is a causal relationship,
the direction of that causality is unknown.
NARRATOR [continued]: When we're running a correlational study, we're measuring at least two
variables and seeing what happens to the value of one as the value of the other construct changes.
First, we want to select two variables that we believe will be related and whose relationship is
important for some theoretical or practical reason.
NARRATOR [continued]: For example, we may be interested in finding out whether the SAT test
actually predicts college aptitude. In this case, we will want to look at both SAT scores and success
in college, perhaps based on grade point averages in the student's freshman year.
NICOLE CAIN: As with all research, you should keep careful track of all of the data that you're
collecting. This includes keeping track of what condition participants are in, keeping track of their
scores on the various variables that you're measuring, as well as any problems or issues that come
up in the course of collecting data. You'll want to have this information available to you when you go
to analyze your results.
NARRATOR: The nuts and bolts of computing correlations requires advanced knowledge and most
people use a statistical software package to accomplish this task. The most common type of
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 9 of 12 Research Design: Observational and Correlational Studies
correlation is the bivariate correlation, which measures the degree of relationship between two
variables. The value that this calculation yields
NARRATOR [continued]: is called the "r statistic." This correlation, r, ranges from negative 1.0 to
positive 1.0. For example, we could look for a correlation between the number of hours students spent
studying and their scores on a math exam by way of these steps.
NARRATOR [continued]: First, we would look for the sign of the correlation, whether it was positive
or negative. A correlation of positive 1.0 would indicate a perfect positive relationship between hours
spent studying and math score. That means that as the number of hours spent studying increases,
the math score also increases.
NARRATOR [continued]: A correlation of negative 1.0 would indicate a perfect negative relationship
between hours spent studying and math score. That means that as the number of hours spent
studying increases, the math score decreases. A correlation of zero would indicate no relationship at
all between the two variables.
NARRATOR [continued]: In reality, correlations are almost never exactly positive 1.0 or negative 1.0.
Variables are usually not that exactly related. Second, we would look at the absolute value of the
number, ignoring the sign, and see whether it is closer to one or to zero. The closer it is to a full
negative or positive 1.0,
NARRATOR [continued]: the stronger the relationship between our two variables. The closer it is to
zero, the weaker the relationship. Let's apply these principles to three sets of numbers. Which is the
stronger correlation, negative 0.98 or positive 0.34?
NARRATOR [continued]: Negative 0.98. Which indicates that both variables increase together,
negative 0.45 or positive 0.21? Positive 0.21. Which indicates that as one variable increases, the
other decreases, positive 0.62 or negative 0.33?
NARRATOR [continued]: Negative 0.33. Once we have calculated our correlation, we need to
understand the implications of that correlation. It is extremely important to be careful at this stage in
order not to misinterpret the results of our correlational study. The most important rule remains that
correlation does not
NARRATOR [continued]: imply causation. This is the biggest mistake one can make when analyzing
data.
EVELYN BEHAR: So let's talk about a specific example. Imagine that you ran a study looking at two
variables that you were interested in, number of ice cream cones sold in New York City and number
of violent crimes committed in New York City. And let's say that you found a very strong correlation
EVELYN BEHAR [continued]: between these two variables. You found that as ice cream cone sales
increased, so did number of violent crimes, that it increased almost at a one to one ratio. Well, we
have to be very, very careful about drawing any causal conclusions here. It could be that ice cream
causes people to be violent.
EVELYN BEHAR [continued]: It doesn't sound very plausible but it is a possibility and the correlational
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 10 of 12 Research Design: Observational and Correlational Studies
study cannot tell you whether it's true or not. It's also possible that when people commit violent crimes,
it makes them somehow want ice cream. This also doesn't really seem to be intuitively true and yet
we cannot rule it out with a correlational design.
EVELYN BEHAR [continued]: It's still a possible hypothesis. The third possibility is that there is some
third variable that's acting on these two variables. So in this example, I think there's a fairly clear third
variable and that's weather. It's what the conditions are like outside. In the summertime, there are lots
EVELYN BEHAR [continued]: of people outside so ice cream sales go up and also number of violent
crimes go up. So remember, anytime you have a correlational study and you find some correlation
between your two variables that you're interested in, there are three possibilities. It could be that
variable A is causing a change in variable B, it could be that variable B is causing a change in variable
A,
EVELYN BEHAR [continued]: or it could be that C, some third variable, is causing both A and B to
change and it's influencing both of them. And again, the only way to distinguish between these three
possibilities and to draw solid conclusions about causality and direction of causality is by running a
true experiment.
NARRATOR: Conclusion. We all think about exciting science experiments in which we act upon one
variable in order to cause a change in another. However, we can also learn a great deal without
experimentation just by observing and taking note
NARRATOR [continued]: of what is happening around us. Let's review the two basic types of non-
experimental research designs. In an observational study, we're seeking to learn as much as we can
about a single construct or variable. This may be a brand new, never studied phenomenon, such as
the example of the neon-green hair.
NARRATOR [continued]: An observational study allows us to formulate hypotheses about a construct.
The advantages of this type of study include its relatively low cost when compared to other types
of research and its application in cases where it would be immoral or impossible to cause a
phenomenon to occur. The disadvantages of observational design
NARRATOR [continued]: include its inability to tell us anything conclusive about the relationship
between one construct and another. Correlational studies enable us to find positive and negative
correlations between variables, to see whether they both increase together or whether one increases
while the other decreases
NARRATOR [continued]: or whether they bear a very weak or nonexistent relationship. Correlational
studies are particularly well-suited in situations in which we cannot conduct an experiment, such as
when it would be unethical to cause one condition in order to see what its result would be. In other
cases, a correlational study is a precursor to an experiment, much as an observational study
NARRATOR [continued]: can be a precursor to either a correlational or an experimental study. It
enables us to formulate hypotheses which we can then research further. A drawback of correlational
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 11 of 12 Research Design: Observational and Correlational Studies
designs, which is shared, of course, by observational designs, is that we cannot use them to
determine whether one phenomenon causes another.
NARRATOR [continued]: Even if we find a strong correlation, we're left with the question of whether
the first influences the second, the second influences the first, or a separate third variable influences
them both. The only way to solve this dilemma is, when possible, through experimental research.
[MUSIC PLAYING]
SAGE 2011 SAGE Publications, Ltd. All Rights Reserved.
SAGE Research Methods Video
Page 12 of 12 Research Design: Observational and Correlational Studies
- Research Design: Observational and Correlational Studies