Chapter 8 finish start ch. 7 class version
Transcript of Chapter 8 finish start ch. 7 class version
Factorial Designs
• So far: basic designs (one IV, one DV)
• Now: more than one IV (still one DV)
2 x 3 Factorial Design
Independent Variable A A1
A2 A3
A1 B1Cell mean
A2 B1 A3 B1
A1 B2 A2 B2 A3 B2IV B
B1
B2
B1Marginalmean
B2Marginalmean
A1Marginal
mean
A3Marginal
mean
A2Marginal
mean
2 types of effects
• Main Effect - The influence of one Independent variable in a factorial design
• Interaction Effect - joint influence of two or more IVs on the DV– The effect of one IV depends on the level of
another IV.
Example
• Study examining gender (M-F) and intervention to improve test-taking skills
• 3 IV levels– control (no intervention)– reading material (instructional booklet)– personalized tutoring
Intervention Control Booklet Tutoring
A1 B1 A2 B1 A3 B1
A1 B2 A2 B2 A3 B2
Male
Female
What would a main effect of gender look like?
Control Reading Tutoring0
102030
405060
708090100
MaleFemale
What would a main effect of intervention look like?
What would an interaction look like?
What should we interpret?
• If one main effect - report it• 2 main effects – report both• BUT if there’s an interaction…
– Only interpret/report the interaction– Because the effect of test-taking intervention
depends on gender
Interaction
Combining Between and Within Participant Designs
• Factorial design based on a mixed model• -or- mixed model design• IVs can be either between-groups (e.g.,
gender) or within-groups (a.k.a. repeated)
Advantages of Factorial Designs
• Can test more than 1 hypothesis at a time • Able to deal with extraneous variables
– Build into design and test outright
• Increases precision b/c it evaluates more variables at once
• Allows researcher to understand interactive effects of variables
Disadvantages of Factorial Designs
• Gets messy with more than 2 IVs
• Requires more participants (N per cell)
• More difficult to simultaneously manipulate all IVs when you have more of them
Choosing an Experimental Design
• Depends on…• Research question• Nature of variables you are investigating• We have discussed design building blocks
• Page 255: guiding questions
Chapter 7
Control Techniques in Experimental Research
Overview
• Control at the beginning of experiment– Random assignment– Matching
• Control during the experiment– Counterbalancing– Controlling for participant effects– Controlling for experimenter effects
Create equivalent experimental groups
Treat groups the same during the experiment
Random Assignment
• Not to be confused with random sampling!
• In reality, random sampling is rarely used in experimental research
• Generalize on the basis of multiple studies• With different kinds of samples/settings
Random Assignment
• a.k.a randomization– Most important of all control methods– Only technique for controlling both known
and unknown extraneous variables
In experimental
design
Random Assignment
• Quiz time:• How does randomization eliminate systematic
bias in experiments (produce control)?
– All variables distributed in approximately the same manner in all groups
– Influence of extraneous variables is held constant
Random Assignment
• Sample Size– It is possible for random assignment to fail – rare with a large enough sample size (N > 30)
Random Assignment
• Ways of achieving randomization– Table of random numbers– Randomizer.org– Draw out of a hat– Be creative – flip a coin/lottery/etc
• www.Randomizer.org
Text pp. 203-207
Matching
• Equate participants on one or more selected variables
• Matching Variable: The extraneous variable used in matching
• Useful when random assignment is not possible
Methods for Matching Participants
• Holding variables constant• Building the extraneous variable into the
design• Yoked control• Equating participants
Matching by Holding Variables Constant
• Hold extraneous variable constant for all groups in the experiment
• All participants in each treatment group will have same degree or type of extraneous variable
• Requires selection criteria for participant sample
Build Extraneous Variable into the Research Design
• Especially useful if you are interested in:– Differences produced by the levels of the
extraneous variable– Interaction between levels of IV and levels of
extraneous variable
• Sound familiar?– What kind of research design would this be?
Example: Effect of a study skills intervention on college grades in a Quantitative Methods
course…Intensive tutoring program Study packets (usual)
But the literature suggests that learning style may affect how students respond to different study skills training methods.
Learning style is a potential confounding extraneous variable….but we can build it in to the design!
Learning Style
Visual Auditory Kinesthetic
Intensive tutoring program
Study packets
Inte
rven
tion
A
B
Matching by Yoked Control
• Match participants on the basis of the sequence of administering an event
• Each control participant is “yoked” to an experimental participant
• Controls for the possible influence of participant-controlled events
• Example: Sklar & Anisan (1979)– stress and immune response
Matching by Equating Participants
Precision control• Match each participant in experimental group
with a participant in control group on variable(s) of concern
• Example: Scholtz (1973) compared defense styles in suicide attempt vs. no attempt
Matching by Equating Participants
• Precision Control Advantage– Groups are equated on matching variables
• Precision Control Disadvantages– How do you know which variables are critical?– Difficulty of finding matched participants increases
exponentially as number of matching variables increases– Matching limits generalizability of results– Some variables are difficult to match
• Example: prior psychotherapy
– Matching can only be as accurate as the available measurement of the matching variable