Académique Documents
Professionnel Documents
Culture Documents
DESIGN*
main source: Vernoy & Vernoy (1997) & Falik & Brown (1983)
How a well
well--controlled and well
well--designed
experiment (research) looks like?
IDENTIFYING VARIABLES
I d
d t Variables
V i bl
Independent
Independent variables are those that
are manipulated by the experimenter
and applied to the subject in order to
determine what effect they may have.
These are the factors that the
experimenter varies from one condition
to another, such as the items in
memory lists or the dosages of an
experimental drug.
In an experiment, a researcher may
want to compare the effects of two
different independent variables or the
effects of varying amounts (such as
different dosages of drugs) or levels
(such as low imagery or high imagery in
a memory experiment) of one particular
independent variable.
D
d tV
i bl
Dependent
Variables
Dependent variables are those that are
used to assess or measure the effects
of the independent variables.
In psychology, dependent variables are
measures of the behavior being
studied
studied.
The dependent variable depends on the
independent variable; if the variables
are causally related, the dependent
variable will vary, or change, as the
independent variable changes. Thus,
dependent variables are measured by
the experimenter to determine whether
the experimental procedures result in
any behavioral change.
Examples: the number of visual images
identified correctly (dependent) and the
amount of time it takes to push a button
(independent).
S
bj t V
i bl
Subject
Variables
Subject variables are those that describe subjects'
characteristics or attributes
attributes, factors that cannot be
manipulated by the experimenter.
Examples of subject variables are gender, IQ,
ethnicity age
ethnicity,
age, etc
etc.
S
bj t V
i bl
Subject
Variables
Subject variables are those that describe subjects'
characteristics or attributes
attributes, factors that cannot be
manipulated by the experimenter.
Examples of subject variables are gender, IQ,
ethnicity age
ethnicity,
age, etc
etc.
P01
P02
P03
P04
P05
P06
P07
P08
P09
P10
P11
P12
P13
P14
P15
x1
x2
x3
x4
x5
x6
x7
x8
x9
x10
x11
x12
x13
x14
x15
y1
y2
y3
y4
y5
y6
y7
y8
y9
y10
y11
y12
y13
y14
y15
where
X
the total amount of lovastatin (in mg) taken by patients (independent variable)
Factors Affecting An
Experimental Design
The experimental design depends on:
F1
the n
number
mber of independent variables
ariables
F2
F3
F4
th samples
the
l ((size,
i
group))
A within
within--subjects design (also
called a repeated-measures
design) requires that each
subject participates in all levels
of all independent variables;
thus, each subject stays within
the experiment for its entire
run.
Between--Subject Design
Between
Each
E
h subject
bj t participates
ti i t in
i only
l one off the
th
experimental conditions and is independent of the
other groups. The subjects in the no-imagery/norelaxation condition are different from the subjects
j
in the imagery/relaxation condition, and so forth. In
this way, subject effects such as innate athletic
ability are randomized throughout all conditions.
A between
between--subjects design
(sometimes called an
independent-group design)
requires that each level of each
independent variable has different
subjects; thus, there is a distinct
difference between each level of
the experiment because each
subject participates in one and
onlyy one level.
Within--Subject Design
Within
Each subject participates in every condition. Thus,
in order to have six subjects in each condition, you
need a total of only six subjects in the entire
experiment.
A within
within--subjects design (also
p
called a repeated-measures
design) requires that each
subject participates in all levels
of all independent variables;
thus, each subject stays within
the experiment
p
for its entire
run.
Mixed Design
A mixed design is used when it is
necessary to have each subject
participate
p
p
in all levels of one
independent variable,
variable while participating
in only one level of another independent
variable. For example, if we were to use
a mixed design
g in the g
golfing
g study,
y, we
could assign a particular subject (S1) to
only the no-relaxation conditions
(Fig.4A). Then this subject participates
solely
y in the no-relaxation/no-imagery
g y
and the no-relaxation/imagery
conditions. Or we could assign the
subject to only the two no-imagery
conditions ((Fig.4B).
g )
Mixed designs use fewer subjects than
between-subjects designs, but they are
y statistically.
y
the most difficult to analyze
Fig. 4
Addressing
One-Group
g F4: Onep
Experimental Designs
The simplest
Th
i l t experimental
i
t l design
d i that
th t iinvolves
l
comparing a single sample mean to the mean of a known
population.
Example:
Suppose you want to know whether the fifth formers in
School X have better IQ than their colleagues in the
country. If you have access to national data for all fifth
formers IQ, you can give the IQ test to all fifth formers
in School X and compare
p
its mean to the mean of the
population.
Fig. 5
The Experiment:
(if you want to test more than two keyboards).
The Experimental Design:
You can certainly test more than two keyboards by adding more groups by
randomly selecting more subjects and assigning them in a random fashion to
the additional groups. But adding more keyboards simply extends the
completely randomized design; it does not increase the number of
independent variables. There is still only one, the type of keyboard (see Fig.
6) You have merely increased the number of levels (number of different
6).
keyboards) of the independent variable.
Fig. 6