Experimental Designs

1
Experimental Designs

• Psychological Research Methods

2
Background

• Purpose
• Research Question
• Hypothesis
• Conclusions
• Contrast with Quasi-Experimental Designs

3
Independent Variable (IV)Types of Manipulation

• Stimuli (e.g., perception studies)
• Context (e.g., competitive vs. cooperative)
• Information (e.g., dating, hiring)
• Martin et al. (1998) and medical diagnosis
• Social setting (e.g., alone or in groups)
• Physiological Conditions
• Stress
• Arousal
• Drugs
• Deprivation (e.g., food, water, sleep, sex)

4
Between Groups Designs

• Different groups
• Group Equivalence
• Confounding Variables
• Definition
• Mask effect of the IV (Type II Error)
• Provide a False Effect (Type I Error)

5
Simple Between Subjects Design
Exp. Group ? Tx ? Test
Random Assignment To Groups
Control Grp. ? ? Test
6
Sources of Variance
7
Methods for Reducing Error Variance

• Holding Extraneous Variables Constant
• Blind and Double Blind Studies
• Weak vs. Strong Manipulation

8
Statistical Analyses

• Ratio and Interval Level Data
• One IV
• 2 levels
• 3 levels
• Two IVs
• Ordinal Level Data
• Mann-Whitney U
• Nominal Level Data
• Chi square goodness-of-fit

9
Within Groups Designs

• Types
• Benefits
• Number of Participants
• Error Variance
• More Powerful Test of Effect

10
Costs Order Effects

• Practice
• Fatigue
• Carryover
• Attrition

11
Controlling Order Effects

• Time between Treatments
• Order as an IV
• Counterbalancing

12
Counterbalancing Techniques

• ABBA
• Complete
• Block Randomization
• Random Order with Rotation
• Latin Square

13
Statistical Analyses

• Interval and Ratio Level Data
• One IV
• 2 levels
• 3 levels
• Two IVs
• Ordinal Level Data
• Wilcoxon signed ranks or Friedmans test
• Nominal Level Data
• McNamars test or Cochrans Q

14
Factorial Designs

• More than One IV
• Components
• Main Effects
• General
• Simple
• Interaction

15
Complete Factorial
16
Incomplete Factorial
17
Simple Factorial Design
Exp. Group ? Pretest ? Tx ? Test
Random Assignment To Groups
Control Grp.? Pretest ? ? Test
18
Solomon Four Group Design
Exp. Group ? Pretest ? Tx ? Test
Exp. Group ? ? Tx ? Test
Random Assignment To Groups
Control Grp.? Pretest ? ? Test
Control Grp.? ? ? Test
19
There are many different methods of instruction
used in the academic setting. Little experimental
research exists examining which mode is more
effective in educating students. The purpose of
this study was to compare scores obtained on the
written and the practical examinations of
students on a single topic taught through either
a teacher-centered format or a student-centered
format. A 2 X 2 X 6 factorial design was used in
this study. Independent variables were teaching
style (teacher-centered instruction and
student-centered instruction), order (first or
second), and learning style (competitive,
collaborative, participant, avoidant, dependent,
and independent). The dependent variables were
the scores obtained on a written and a practical
examination of gait and crutch fitting. Forty
pre-athletic training students in their first
semester of their first year (16 males, 24
females) participated in this study. The
Grasha-Reichmann Student Learning Style Scale.
(Livecchi et al., 2004)
20
Illness scripts are knowledge structures
composed of consequences, enabling conditions,
and faults. The effects of illness script
components--consequences and enabling
conditions--and physician factors on referral
decisions for gastrointestinal disorders were
investigated. The hypothesis that consequences
and enabling conditions increase the likelihood
of referral was confirmed and several
interactions between consequences and enabling
conditions were found. The hypothesis that
physician factors moderate the effect of enabling
conditions was also confirmed, but (contrary to
illness script theory) evidence was also found
for moderation of consequences. Both enabling
conditions and consequences were found to be
moderated by physician factors to a larger extent
than previously assumed by illness script
theory. (van Schaik et al., 2005)
21
Although binge eating disorder is. This study
was designed to examine the added benefit of two
adjunctive interventions, individual cognitive
behavioral therapy (CBT) and fluoxetine, offered
in the context of group behavioral weight control
treatment. One hundred sixteen overweight/obese
women and men with binge eating disorder were all
assigned to receive a 16-session group behavioral
weight control treatment over 20 weeks.
Simultaneously, subjects were randomly assigned
to receive CBTfluoxetine, CBTplacebo,
fluoxetine, or placebo in a two-by-two factorial
design. Outcome measures, assessed at the end of
the 16-session acute treatment phase, included
binge frequency, weight, and measures of
eating-related and general psychopathology.
Overall, subjects showed substantial improvement
in binge eating and both general and
eating-related psychopathology. (Devlin et al.,
2005)
22
An experiment was designed to test behavioural
differences in the detection of deception arising
from investigative interviews conducted in either
a first or second language. A two (Cantonese or
English) by two (deception or truthfulness)
between-subjects factorial design was used.
Twenty-six postgraduate criminology student
observers provided judgments of lying in 20
video-taped interviews of undergraduate subjects
randomly assigned to either telling the truth or
lying about their opinions on capital punishment.
Observers did less well in identifying liars in
their first language but were more successful in
identifying liars speaking in a second language.
However, observers made more mistakes with those
telling the truth in a second language. The
degree that deceivers deployed countermeasures
also varied with second language users reporting
less ability to control verbal and non-verbal
behavioural cues. Deceivers. (Cheng
Broadhurst, 2005)
23
The main objective of this study was to
determine the association between cognitive
restraint, nutritional intake and eating patterns
in free-leaving subjects. We administered a
questionnaire that included information about
eating behaviour (restraint, disinhibition, and
hunger), dietary intake, and physical activity
380 students (60 female) completed the study.
The association of restraint (low/high) and
disinhibition (low/high) with dietary intake was
tested using MANOVA in a 22 factorial design.
Statistically significant main effects were
further analyzed using ANOVAs. To identify eating
patterns, factorial analysis was employed. Among
women, high restrainers reported lower
consumption of energy, pastry, and starchy foods,
and higher consumption of vegetables, and fish,
than low restrainers. In male subjects, high
restrainers consumed significantly more
vegetables than low restrainers. (Moreira et
al., 2005)