PP 8 - Personal Web Pages
Download
Report
Transcript PP 8 - Personal Web Pages
Repeated Measures ANOVA
factorial within-subjects designs
One-Way Repeated Measures
One-way Repeated Measures Designs are used
when:
– the same subjects are measured on 3 or more
occasions (TIME)
– the same subjects are exposed to 3 or more
treatments (TREATMENT)
– the same subjects provide three or more ratings
that are measured on the same scale (MEASURE)
Examples
The same subjects are assessed on pre,
mid, and post treatment occasions.
The same subjects are given three
different types of medication.
The same subjects rate three different
aspects of school climate.
Factorial Designs
Between-subjects terms can be completely
crossed with within-subjects terms to form
factorial designs.
All three uses of within-subjects terms, TIME,
TREATMENT, and MEASURE, can be
combined with between-subjects terms to form
a variety of completely crossed factorial
designs.
Examples - TIME
Subjects are assessed on pre, mid,
and post treatment occasions,
AND are randomly assigned to two
different treatments.
Pre
Textbook I
Textbook II
Mid
Post
Examples - TREATMENT
Male and female participants each receive
three different treatment conditions.
Participants are randomly assigned to
receive the treatments in different orders.
Drug I
Male
Female
Drug II
Placebo
Examples - MEASURE
Teachers rate three different aspects of
school climate, AND are randomly
assigned to a treatment or control group.
The treatment group gets a particular
model of administrator support.
Communication
Treatment
Control
Support
Policu Clarity
Examples - MEASURE
Subjects
are randomly assigned to
three different types of medication,
AND asked to rate two different
aspects of the effects of the drug.
Pain
Drug A
Drug B
Placebo
Swelling
Educational Evaluation
Factorial designs with multiple
completely crossed within-subjects terms
can also be used but are relatively rare in
educational research.
“Split plot” designs are very common,
with one within-subjects term (time) and
one between-subjects term (group). Why?
Examples
Suppose you are charged with evaluating
different delivery models for staff
development in your school district.
The question is whether some use of
computer based instruction would be
helpful.
Examples
You are interested in evaluating knowledge, job
satisfaction, and teaching effectiveness gains over
time.
You consider the following three specific delivery
models for staff development in your school district:
– Traditional format
– Computer-based tutorials
– The combination of the two
Examples
A possible research design:
Pre
Post
Computer Tutorial
Traditional Delivery
Trad + Tutorial
What are some of the potential issues with
this design?
Examples
What are some of the potential issues with
this design?
–
–
–
–
Randomization of schools, teachers, classes, etc.
Difficulty of content
Time of the year the instruction takes place
Availability of computer technology
Pre
Computer Tutorial
Traditional Delivery
Trad + Tutorial
Post
Our Research Design
Fall
Winter
Spring
48-50
Child
Age in
Months
51-53
at First
Assessment
54-56
Outcomes:
Social-Emotional
Literacy
Cognitive
Language
Physical
Mathematics
Factorial Designs
Just like the One-way ANOVA is
analogous to the One-Way
Repeated Measures procedure,
Split plot factorial designs share
many of the same properties with
completely crossed BetweenSubjects Factorial designs.
Similarities
Null and Alternative Hypotheses for Multiple
Main Effects
Null and Alternative Hypotheses for
Interaction Terms
Graphing the data and post-hoc comparisons
are essential as interpretation aids.
Hypotheses
Main Effect for Time (MEASURE or
Main Effect for Group
TREATMENT)
Interaction Effects – different patterns of
growth or rates of growth between the groups
Our Research Design
Fall
Winter
Spring
48-50
Child
Age in
Months
51-53
at First
Assessment
54-56
Outcomes:
Social-Emotional
Literacy
Cognitive
Language
Physical
Mathematics
Differences
Sphericity Assumption with the Univariate
case.
Homogeneity of Variance-Covariance
Matrices in the Multivariate case.
Data from individual subjects occurs in
multiple cells rather than only one cell.
Special Considerations
Additional potential threats to the
validity of this type of design:
–
–
–
–
practice effects
order effects
fatigue effects
carry-over effects
Interpretation
Follow the same steps we used for
factorial designs with only betweensubjects terms
Consider the interactions first
Graph the results
Look at Height, Slope, Parallelism
Use Tukey Post Hoc test to help explain
the results
Interpretation
Height = difference between
groups
Slope = growth over time
Parallelism = differential rates of
growth between the groups
Graphs
Social Development by Schedule
4.000
3.500
3.000
2.500
2.000
1.500
Split Day
Fall
Winter
Spring
Split Week