This document discusses different research design strategies. Weak designs do not control for extraneous variables and have threats to internal validity. Strong designs use control groups and techniques to eliminate rival hypotheses. Control groups allow comparison to estimate the effect of a treatment by controlling for other influences. Random assignment creates equivalent groups and controls for threats to internal validity. Pretest-posttest designs with control groups allow comparison of change scores and ensure initial group equivalence.
2. Research Design
The plan or strategy used to answer the
research question
This is where we put to use all of the things
that we have learned about validity and
reliability.
4. Weak Experimental Designs
Do not control for many extraneous variables
Provide weak evidence of cause and effect
Have threats to internal validity
5. One-group Posttest- Only
Design
Influence of a treatment is investigated with
only one group of individuals
Single test after treatment (“posttest”)
Example: New reading program in 1st grade
Does program improve reading skills?
7. Problems?
Extraneous variables?
Why/when would you use this design?
If there is background information on DV and prior
work has shown mechanism by which IV impact DV.
8. One-group Pretest-Posttest Design
A treatment condition is interjected between
pre- and posttest of the dependent variable.
Pretest measure Treatment Posttest Measure
O X O
Compare
9. One-group Pretest-Posttest Design
Advantage over one-group posttest?
Problems?
Threats to internal validity?
Why would we use this design?
If we cannot create an equal group for comparison, we can
get SOME information this way
10. Nonequivalent Posttest-Only
Design
Performance of an experimental group is
compared with that of a nonequivalent control
group at posttest
Posttest
Treatment Measure
Experimental Group X O Compare
Control Group O
11. Advantage over previous designs?
Problems?
Threat to internal validity?
Selection
Best thing to do is to use random assignment.
Creates equivalent groups
Next best is to use matching on relevant variables.
Will discuss this next chapter
13. To control for extraneous variables you must
eliminate potential rival hypotheses.
Done in two ways:
Control Techniques- more on this next week (ch. 7)
Control Group
14. Control group
comparison group
Does not receive “active” level of IV
○ No intervention or standard experience
Experimental group
receives a level of the IV intended to produce
effect
15. Functions of a Control Group
1. Serve as a comparison to the experimental
group. Did the treatment/manipulation have
an effect?
Estimate Counterfactual- what the participants’
responses would have been had they not
received the treatment
16. Functions of a Control Group
2. Control for rival hypotheses
With random assignment, extraneous variables
will have an equivalent impact on both groups
17. Strong Experimental Research
Designs
Basic designs – one IV and one DV
Between-participants
Within-Participants (repeated measures)
Factorial Designs – multiple IVs
18. Posttest-Only Control Group
Design
This design looks familiar, right?
What is different now?
Posttest
Treatment Measure
Experimental Group X O Compare
Control Group O
19. Posttest-Only Control Group
Design
We could have more than 1 experimental
group
Posttest
Treatment Measure
Control Group O
Experimental Group 1 X1 O Compare
Experimental Group 2 X2 O
21. Pretest-Posttest Control Group
Design
Simply add pretest to previous design
What comparisons will we make?
Pretest Posttest
Treatment
Measure Measure
Experimental
O X O
Group
Control
O O
Group
22. Benefits of Pretest
Ensure equivalency of groups
Detect ceiling and floor effects
Select participants accordingly
Can empirically demonstrate effect of treatment
Change scores
See if initial position on DV is important
Treatment may affect hi/low differently