Main Effects

1
Main Effects Interactions
2
Consider the following hypothetical study
--we are interested in evaluating a new treatment
for aggression.
--we recruit 20 children, 10 boys and 10 girls
from one school
--we randomly assign half of the boys to a
therapy group the other half to a control
group, and we do the same with the girls
--we give the 5 boys and 5 girls a group-based
therapy for aggression
--we give the 5 boys and 5 girls in the control
group a group-based game playing task that has
nothing to do with aggression
--after one month we evaluate all of the children
on a 1 to 5 scale of aggressive behavior in the
classroom
3
In what research questions might we be interested?
--ignoring treatment condition, are there gender
differences in aggressive behavior? This is the
main effect of gender.
--ignoring gender, is there evidence that the
therapy condition decreases aggressive behavior
compared to the control condition? This is the
main effect of treatment.
--do boys and girls respond differently to the
same therapy? This is the interaction between
gender and treatment.
--when there are differences among the levels of
a single factor, ignoring the other factor, we
say there is a main effect of that factor
--The term interaction means that factor 2
affects the different levels of factor 1
differently, or vice versa. For this the therapy
affects boys and girls differently. It depends
is another way of describing an interaction the
effect of factor 1 depends on the level of factor
2.
4
A main effect of gender, ignoring treatment
5
A main effect of treatment, ignoring gender
6
A complete lack of effects of gender or
treatment no interaction
7
A main effect of gender, no effect of treatment
no interaction
8
A main effect of treatment, no main effect of
gender, and no interaction between treatment
gender
9
A main effect of treatment, a main effect of
gender, and no interaction between treatment
gender
10
An interaction between treatment and gender main
effect of treatment no main effect of gender
11
An interaction between treatment and gender no
main effect of treatment or gender
12
More on interactions
Interaction of A and B If the effect of one
factor depends on the different levels of a
second factor, then there is an interaction
between the factors
See also this supplemental handout on interactions
13
(9)
Differences in the differences
14
3-way interactions

• Now lets say youd like to compare two different
  age groups of boys and girls on treatment for
  aggression.
• Perhaps the difference in aggression between boys
  and girls is minimized in middle school, and
  maximized in adulthood. If you administer a
  treatment for aggression to both genders at both
  age groups, and find that the interaction between
  treatment and gender is different at adolescence
  and at adulthood, then you have a 3-way
  interaction.

15
Is there a 3-way interaction here?
16
Variables with more than 2 levels

• A researcher wants to determine the most
  effective treatments for clinical anxiety
  disorders. The most common anxiety disorders can
  be classified as either simple phobias or
  complex phobias. The researcher plans to
  compare anxiety scores of patients diagnosed with
  some type of phobia after they have received
  either systematic desensitization (behavioral
  therapy), cognitive-emotional therapy, or a
  placebo treatment.

17
(No Transcript)
18
Project 2

• According to DeNeys (2006), we make decisions in
  two ways
• Relying on knowledge and beliefs (also called a
  heuristic system)
• Is considered effortless, rapid, and automatic
• Relying on logical standards (also called an
  analytic system)
• Is effortful, slow, and makes high demands on
  computational resources.
• Therefore, the more resources that are available,
  the more likely youll be able to engage this
  system

19
DeNeys (2006), cont

• Sometimes, these two processes conflict.
• Lets say youre given the following syllogism,
  and your task is to evaluate whether the
  conclusion follows validly from the first two
  statements
• All mammals can walk. Whales are mammals.
  Therefore, whales can walk.

20
DeNeys (2006), cant

• The idea that whales can walk clearly conflicts
  with your knowledge about the world. But, if you
  have the computational resources available, you
  will be able to rely on your logical standards
  and reason that it still follows validly from the
  first two assumptions.
• But what if you do not have the computational
  resources available?
• For instance, what if at the same time that you
  were asked to evaluate the syllogism, you were
  also instructed to remember complicated patterns?
  
• DeNeys measured error rate on the syllogisms as a
  consequence of concurrent cognitive demand, and
  he later indicated that looking at reaction time
  may be an additional consequence to investigate.

21
Furthermore

• Whether someone has a high or low working memory
  capacity may moderate the effect of cognitive
  load on error rate.
• How? What do you expect to change?
• Think in steps
• How would working memory capacity affect response
  on the easy/automatic questions?
• How would it affect response on the
  difficult/resource-demanding questions?
• How would it affect cognitive loads effect on
  the easy questions?
• How would it affect cognitive loads effect on
  the difficult questions?

22
Project 2 goals

• For project 2, we conducted an experiment very
  similar to that used by DeNeys (2006). We are
  ultimately interested in the following questions
• Is performance on the reasoning task related to
  individual difference measures of working
  memory?
• Does cognitive load affect error rate? What
  about reaction time?
• Does the effect of cognitive load on error
  rate/reaction time depend on a subject's working
  memory capacity?
• Is there evidence for two separate reasoning
  processes?

23
Project 2 Next steps

• Come to class on Tuesday (3/6) having read DeNeys
  (2006), having reviewed the slides on
  interactions, and with a strong understanding of
  the goals and predictions of project 2
• On Tuesday, well look at the data and do
  analyses. Your ability to interpret the data
  relies entirely on your understanding of
  interactions and of what differences we expect to
  find, so please come prepared.