Transcript PPT Lecture Notes

Introduction
To
Complex Designs
Outline of Today’s Discussion
1. Introduction to Complex Designs
2. Analysis of Complex Designs
Part 1
Introduction To Complex Designs
Intro To Complex Designs
1.
According to our text, what is a “complex design”?
2.
Potential Pop Quiz Question: What is a synonym for
“complex design”?
3.
Potential Pop Quiz Question: What is the simplest
complex design?
4.
Potential Pop Quiz Question: Explain the primary
advantage associated with complex designs?
Intro To Complex Designs
1.
Let’s get some practice reading graphs.
2.
Some of the following graphs are from complex
designs, and some are not.
3.
In addition to describing each graph, indicate whether
(and why) the design is simple or complex…
Intro To Complex Designs
Describe how the variables are plotted in this graph.
How many “levels” does the IV have, and what are they?
Intro To Complex Designs
Describe how the variables are plotted in this graph.
How many IV’s, and how many levels of each?
Intro To Complex Designs
Describe how the variables are plotted in this graph.
How many IV’s, and how many levels of each?
Intro To Complex Designs
Here’s an oldie but goodie…
Which graph is “messed up”?
Part 2
Analysis of Complex Designs
Analysis of Complex Designs
1.
In this section, we’ll focus on three concepts, main
effects, interactions, and simple effects.
2.
We’ll consider each in turn…
3.
Main Effect - The overall effect of one independent
variable in a complex design.
4.
A main effect of one IV is assessed by “collapsing
across” (averaging over) all levels of the other IV.
Analysis of Complex Designs
1.
When we consider main effects, we temporarily simplify the
research design.
2.
For example, in an AxB design (read, “A by B”), we might have 2
levels of factor A, and 3 levels of factor B.
3.
When we consider the main effect of factor A, factor B “goes away”
temporarily, and we’re back to a simple design with one factor that
has 2 levels.
4.
When we consider the main effect of factor B, factor A “goes away”
temporarily, and we’re back to a simple design with one factor that
has 3 levels.
Analysis of Complex Designs
Would someone try to describe the main effect
of training (i.e., factor A) here?
Analysis of Complex Designs
Would someone try to describe the main effect
of task (i.e., factor B) here?
Analysis of Complex Designs
1.
Now, let’s develop some intuitions about interactions…
2.
Can someone think of two chemicals that are harmless
when alone, but lethal when they “interact”? This is only
a mediocre analogy. Let’s define it.
3.
Interaction - the phenomenon that occurs when the effect
of one IV depends on the level of a second IV.
Analysis of Complex Designs
1.
Good scientists can detect interactions using a couple of
tricks. (Very sneaky).
2.
The first trick requires looking for parallel lines (or at least
parallel trends) in graphs.
3.
If the lines or trends in a graph are parallel, then there is
no interaction.
4.
To the extent (note the caution in “to the extent”) that lines
or trends depart from parallel, there is an interaction…
Analysis of Complex Designs
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
How cool is this?
Analysis of Complex Designs
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
Is there a unique character here?
Analysis of Complex Designs
Is there a unique character here?
Analysis of Complex Designs
From Shaughnessy, Zechmeister & Zechmeister
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
From Shaughnessy, Zechmeister & Zechmeister
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
From Shaughnessy, Zechmeister & Zechmeister
What is the “dimensionality” of this experiment?
Is there an interaction here?
Analysis of Complex Designs
1.
A second trick requires a little more effort, and is less
intuitive…but it still works…
2.
The subtraction method for finding interactions involves
comparing the differences between the means in each
row (or column) of a means-table.
3.
To the extent (note the caution in “to the extent”) that
the differences are different, there is evidence for an
interaction….
Analysis of Complex Designs
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
Is there an interaction here?
Also, could someone describe the main effects?
Analysis of Complex Designs
1.
Questions so far on interactions?
2.
Let’s now get just a little more complicated, and move to a
complex design that has three independent variables.
3.
We’ll consider the simplest 3-way design, i.e., a 2x2x2.
4.
The first problem we encounter is that we draw graphs on
2-dimensional surfaces…but we have 3 dimensions of
variability in our independent variables.
5.
Any suggestions?
Analysis of Complex Designs
From Shaughnessy, Zechmeister & Zechmeister
Is there an interaction here?
3-Way Interaction
Vertical Activity
Vertical Activity Day 1
Sig 2-Way Interaction
stress
no stress
1400
1200
1200
1000
1000
800
600
400
800
600
400
200
200
0
0
saline
drug
drug treatment
stress
no stress
1400
vertical Activity
vertical activity
Vertical Activity Day 2
No 2-Way Interaction
drug
saline
drug treatment
Denison Student Research on Cross-sensitization to Amphetamine By Mandy Calkins (2004) and Professor Susan Kennedy
Analysis of Complex Designs
1.
For a 2-factor experiment (AxB), we can have a main
effect of A, a main effect of B, and an AxB interaction.
2.
For a 3-factor experiment (AxBxC), we can have a main
effect for each of the variables, but many different
interactions.
3.
The interactions could be AxB, AxC, BxC, or AxBxC!
Things get complicated quickly. :-)
Analysis of Complex Designs
1.
Let’s return to a humble, AxB design and consider a few
additional points about interactions.
2.
First, we should be careful about reporting interactions if
we have reason to believe there may be floor or ceiling
effects…
Analysis of Complex Designs
From Shaughnessy, Zechmeister & Zechmeister
In your own words,
why is this “interaction” bogus?
Analysis of Complex Designs
1.
Also, it is good practice to look at main effects only after
establishing that the interaction is non-significant.
2.
In other words, a significant interaction can complicate
the interpretation of main effects…
Analysis of Complex Designs
Probability of
Finding Spouse
Interactions Obscure
Main Effects
0.8
0.6
0.4
0.2
0
Young
Old
Male
Female
Gender
Here, it could be misleading to say
that there is no main effect of gender or age.
The interaction can “obscure” a main effect.
Analysis of Complex Designs
Probability of
Finding Spouse
Interactions Obscure
Main Effects
0.8
0.6
0.4
0.2
0
Young
Old
Male
Female
Gender
What’s “wrong” with this graph?
Analysis of Complex Designs
Probability of
Finding Spouse
Interactions Obscure
Main Effects
0.8
0.6
0.4
0.2
0
Young
Old
Male
Female
Gender
Shaughnessy,JJ, Shaughnessy, EB, and Zechmesiter, JS.
Research Methods
Psychology,
Hill.
What’sin“wrong”
withMcGraw
this graph?