Ad:
Inferential experiments are considered the most
important method for establishing evidence or proof of causation. This is
because the inferential experiment is really a situation a scientist creates
that simulates the conditions under which the causal influence is supposed to
occur. In the experimental situation, a causal effect is allowed to occur or to
not occur. If it does occur, it is evidence that the causal link really exists.
If it does not, the seeming causal link may not be real.
In an inferential experiment, the experimenter
makes a formal statement of the causal relationship that the experiment is
designed to prove or to disprove. That statement is call the
hypothesis.
For example, let's say we wanted to find out if a
particular antibiotic will kill a particular germ in an infected patient. We
could find a hospital that has patients that have the infection, and give them
the antibiotic. Then either the antibiotic causes the germs to die and the
infection gets better, or it does not, right? Consider this however ... what if
the germs die in the patients that got the pill, but in that hospital it also
dies in the patients that didn't get the pill? You see, we have to look at
the other side of the coin. We could have spend millions making an
antibiotic that didn't really work when really it was something else in the
hospital that killed the germ.
Now, let's say we want to find out if a certain
drug can cause depression patients to become happier, more active, and less
depressed. Well, we could round up a bunch of depression patients and give them
the drug, right? (The people we are experimenting on are called the subjects
or the participants.) If they become more active how
will we know, if the drug really worked or if they just saw a bunch of other
people there and started a party. To make sure we would have to conduct the
experiment under the right conditions.
The basic idea of an experiment is to treat two
groups of participants exactly the same For example, lets say that we get depression
patients together again, but we only give some of them the pill. The ones
who get the pill are in the experimental group because they are
the ones we are REALLY experimenting on - heh heh! After giving the experimental
group the pill we later notice that the ones who got the pill started partying
while the other one's seem to stay depressed. We call the other ones the
control group because they provide a basis for controlling the
influence of other unknown factors in the experimental setting we may not know
about, like the other factors in the hospital that killed the germs in
the first experiment I talked about above. The overall treatment give to the
control group is the control treatment.
Meanwhile back at the lab ... your lab assistant
is in charge picking patients for the experimental group and the control group.
He notices that about half the patients are dressed as clowns (a home remedy for
depression) so he tells the patients dressed as clowns to get in the
experimental group line for one pill and the ones not dressed as clowns should
get in the other (control group) line. Do you see a problem here?
This is a kind of experimenter bias.
You are selecting one group based on some criteria (being dressed as a clown)
that could reflect a difference in their depression level (maybe the people
dressed as clowns are more cheerful, since perhaps its hard to feel depressed
while wearing a big red nose, or maybe they are more depressed, after all who
likes to be referred to as that clown.) Anyway by selecting this
way you may introduce a difference between the group that is not due to the
medication. After all, you want to know if the medication makes a difference,
not the big red noses.
How do you select people for each group then? Flip
a coin? YES, well perhaps you would not literally flip a coin (you would be more
likely to use a computer program) but you would assign participants to the
groups randomly. Any participant could be assigned to either group, which group
he or she ends up in is literally just the luck of the draw.
Why random? Because random it is completely
unbiased, you can't influence who gets in what group so you can't accidentally
introduce some difference between the groups. This is how it is done in a proper
experiment.
OK, so lets say you assign the participants to
each group randomly, and you give the experimental group the pill. Well, if the
experimental group starts partying, and the control group doesn't, it must mean
the pill worked, right? Wait a minute, what if the experimental group started
partying just because they were the lucky ones who got the pill, and the control
group stayed depressed because they didn't get anything? Hey, this kind of thing
used to happen so often that there is a special name for it, its called the
placebo effect. The placebo effect is an improvement that happens
just because the participant got some kind of something, a pill or some other
kind of treatment. This placebo effect can mess up the results of your
experiment by making you think that the medical action of your pill made the
people better when really it was just the fact that they think somebody or some
thing is helping them.
The way this effect, the placebo effect, is
usually controlled is by giving both the experimental group and the control
group something. For example, where the treatment is a pill, you
would give the experimental group the medication and the control group a
harmless sugar pill made to look like the medication. That way both groups get
the feeling that they are getting something. When you give the control group
their something, that something is called a placebo. OK, so in an experiment
like this everybody has to get something so that the placebo effect is the same
in both the experimental and control groups. Again, we want the experimental and
control groups treated exactly the same, except for that one thing ...
in
every detail except one. That one factor is the experimental
treatment that we give only to the experimental group. Later
we examine the two groups to see if the experimental treatment made a difference
between them. If there is a difference then the cause must have been the
experimental treatment.
Now that we have discussed all this placebo stuff,
let's say that you do the experiment again, only this time the experimental
group gets the experimental treatment and the control group gets a placebo. Then
you notice one of your lab assistants over with a group of patients telling
jokes and slapping patients on the back. Then you notice in the record that
these are the patients in the experimental group. YIKES!! What if your lab
assistants start cheering up the experimental group and ignoring the control
group, could that mess up the results??
This sort of conduct on the part of the
researchers in an experiment is also form of experimenter bias. If
you call the research assistant over and you say "don't you see you are messing
up the experiment, it's the pill that is supposed to cheer up the patients, not
the experimenters." But your assistant says, "Gee doc, I was just cutting up a
little bit, you know, ... I'm happy for them getting better and all and besides,
they all knew that they were the one who got the REAL pills by the way Lois was
grinning at them when she gave them the pills." "Lois just looked guilty when
she handed out the fake pills, oh, you know those placebo pills, I think she
just felt sorry for them ... heck I don't blame her and ... (blah blah blah)."
Do you see a problem here?
Yeah, this kind of thing does bias the results of
experiments because, as I said before, its the medication, and not the
experimenters, that's supposed to make the difference. There is a special
procedure for preventing this kind of bias. It is called the double blind
procedure.
Why is it called double blind and not just plain
blind? Its called double blind because there are two groups who are
blind to who gets what pill. These two groups are the participants
themselves, and (guess who else) ... those kind and friendly but rascally lab
assistants. That's right! If the lab assistants (or any other member of the
research team who has contact with the patients) knows who got what pill, and
they could start treating the participants differently, that difference in
treatment could account for final differences between the two groups.
During the double blind procedure, there is
usually just one researcher who puts the pills in bottles and other researchers
(who don't know which pill is in which bottle) actually give the pills to the
participants. In an experiment like this there is usually a special list of
questions that is used by a researcher to measure just HOW depressed each
participant is. A member of the research team fills in the answers while
interviewing the participant, and records the results. You especially want to
make sure that the researcher who administers any such measurement does NOT know
whether the participant he or she is interviewing in is the experimental group
or the control group. If they did know, it could bias the way they perceive and
react to the participant, thereby resulting in biased results.
Now let me explain what is meant by the terms
independent variable and dependent variable. As you
already know, a variable is a concept that can take on more than
one value (in other word it can vary). For example, depression is a concept that
can take on the values of extreme, severe, moderate, mild, none, and perhaps
elated could be another value. Another possible variable, for example is
intelligence, which could be below average, average, above average, etc.
In the experiment we talked about above you have
one variable ( like a pill) that is supposed to exert a causal influence on
another variable, depression. Well the variable that is supposed to be causally
influenced is called the dependent variable, because (using the
experiment we discussed above as an example) how depressed you are is supposed
to depend on whether you got the medication.
The independent variable is the something that is
supposed to exert the causal influence. Again, using the experiment we
discussed above as an example, the independent variable is the
medication. You might say, if it's the medication what are the
values it can take on? In the experiment above it take on the values of present,
as in the real pill, or not present, as in the placebo. Thus you could say that
the independent variable, namely the presence or absence of the medication, is
supposed to exert a causal influence of reducing the level of the dependent
variable, depression.
Students ask me, "why call it the
independent variable ... what's it supposed to be independent of?" No one
has an exact historical explanation. For each participant, the value of the
dependent variable (how depressed they are) is finally supposed to
depend on the value of the independent variable he or she got
(medication present or not present). But whether or not a particular participant
got the medication must NOT depend on how depressed he or she was. In fact, you
are specifically not supposed to let someone's level on the dependent variable
influence what they get on the independent variable.
OK, here are the last concepts in experimentation
that I need to mention. In the experiment discussed above, the difference made
by the independent variable is the difference between the two
groups. You can tell if the independent variable made a difference or
not by looking at the difference between the two group on the
dependent variable. This is called a between groups
experiment.
What if we couldn't get enough participants? Well
we could try giving the pills to all the participants, but at different times.
Let's say in the first week of the experiment we give some of them the real pill
and some the placebo, and the second week we reverse it so the ones who got the
real pill the first week get the placebo the second and the one who got the
placebo the first week get the real pill the second. We would want to see if
there is a difference in each individual's depression when he or she was getting
the real pill versus when he or she was getting the placebo. The difference we
are looking for here is not the difference between two different groups, but the
difference within the participants at different times. This is
called a within group or within groups experiment.
In a within groups experiment there is no separate control and experimental
groups, but the terms experimental treatment and control treatment are still
used.
When I started this discussion I said that I was
going to discuss a simple or classic inferential experiment. Such an experiment
has just one control group and just one experimental group, but in real
experimentation you may have numerous experimental and control groups. For
example, you could have one control group who gets no placebo, one who gets a
placebo, and another one who gets an old treatment for depression (if you are
trying to prove a new treatment is better). You could have an experimental group
who gets the pill, another who gets psycho-therapy instead, and a third
experimental group that gets both pill and psychotherapy.
The reason I also called the simple experiment
classic is because it includes all the basic terms you need to understand
in order to get a good grounding in what experimentation is about. I also
mentioned the within groups experiment because it is one of the most common
types of experiments. Within groups experiments are appropriate for a lot of
different situations, they use is not limited to situations where "you don't
have enough participants".
Lastly, you need to remember that all inferential
experiments do not involve pills or drugs. You could, for example, experiment
with the effectiveness of training the participants a learning strategy, and see
if those who are taught the strategy (the independent variable) learn to perform
a particular task (dependent variable) better than another (control) group who
didn't get the learning strategy. The possibilities for different kinds of
experiments are endless. By understanding the concepts in experimentation
presented here, you will be better able to evaluate and understand the thinking
that goes into planning and evaluating an experiment.
Here are some key terms again:
independent variable - in an experiment the factor
that which is suppose to make "it" happen, the causative factor.
dependent variable - the variable that is
supposed to be affected or changed by the independent variable, so its value depends on the value of the independent varialbe.
hypothesis - a statement of what the
experiment is supposed to prove. control group - in an experiment, the group that does
not get the experimental treatment, they provide an "untreated" basis of
comparison for the experimental group
experimental group -- in an experiment, the group
that does get the experimental treatment. If the independent variable had its
supposed effect, that effect will be reflected as the difference between the
control group and experimental group
primary research - research in which data is actually
collected from the natural world (including experiments, naturalistic
observation, case studies, etc.). This contrasts with secondary research that
draws information from books or publications or expert opinion.
"between groups" vs. "within groups" design - in an
experiment, a "between groups" design has at least one experimental group and
one control group. The effect of the independent variable (the experimental
treatment) is measured by examining the difference between the control group and
the experimental group on the dependent variable. Thus the effect of the
experimental treatment measured as the difference between the groups on the
dependent variable. In a "within groups" design, the same group receives the
control treatment at one point in time and the experimental treatment a another
point in time. The effect of the independent variable is measured by examining
the difference in the dependent variable at the time of the the control
treatment and the dependent variable at the time of the experimental
treatment.
placebo - in an experiment, the control treatment in
which the control group is treated in a harmless but unhelpful way, without know
that the treatment is really not meant to elicit a change in the dependent
variable. An example placebo is a sugar pill which resembles the real medication
used in the experiment.
experimenter bias - any effect that the behavior of
the experimenter has on the experiment that leads to erroneous results.
double blind procedure - a procedure used in an
experiment where some of the experimental team and all of the participants are
not informed as to who is in the experimental group and who is in the control
group. For the experimental team, those who actually administer a drug and those
who take measurements on the participant are especially kept in the dark
concerning who is in which group.
Example:
Aspirin helps to relieve common headaches.
Example: Training in temper control can help people with
anger control problems to be more successful on the job.
Find Your Accredited Online Graduate School
Find Your Accredited Online Psychology Graduate Degree Program
Find Your ONLINE Bachelor's or Associate's College Degree
Arrive2.net: Websites of Bernard Bart Schuster
Free the Palm Trees: Free Palm Tree Clipart
Free Nature Clipart
My Twitter Stream: Twitter.com/arrive2_net.
Words of Wisdom and Inspirational Pictures
Thanks for visiting my web site. Your questions or comments
are welcome!! You can reach me, Bernard Schuster, by e-mail at: Colorado93@hotmail.com .
Squeaker says...
The content of this page is Copyright 2000, 2001, 2012 Bernard Schuster. The term WebQuest,
and template design are Bernie Dodge's.
Teachers can print copies of the pages to use as teaching aids for their
classes, email me for permission before installing electronic copies.
This page provides explanations of common terms used in
experiments. The intent is help students to develop their understanding of the
basic concepts in experimentation. Ad Choices content is selected by Google.
... "see you in the lab!"