EDF5402, Summer 2002
Assignment 5
Due 6-14-02 in class
Part I
The table below is a hypothetical data set, where a group of experimenters investigated the effect of different
levels of anxiety on memory task performance of college students. Assume that eac

EDF5402, Summer 2002
Assignment 4
Due 6-07-02 in class
Use the data in question 7 in the Review Exercises of Chapter 9 in Kirk. Use = 0.05. For the purpose of
hints in questions 9 and 10, label the dependent variable as score, and label the two treatment

EDF 5402, Summer 2002
Assignment 1
Due 5-15-02 in class
Suppose a researcher was interested in the effect of sleep deprivation on the ability of subjects to perform
a vigilance task, such as locating objects moving on a radar screen. The researcher arrang

EDF 5402
Summer, 2002
Experiment Plan 1
Friday, May 17
Think what kind of experiment or quasi-experiment you can conduct in your field of
study. Design an experiment or quasi-experiment with one treatment factor (3 or more
treatment groups) and one depend

Analysis of Covariance (ANCOVA)
A covariate is included into the ANOVA design, in addition
to the dependent and independent variables.
A covariate in an experiment represents a variable that is
known to affect the dependent variable, but not controlled

Split Plot Factorial Design (SPF-pq)
(Ch.12, Kirk)
2 treatment factors (usually fixed effects)
n blocks
The set up is similar to RBF-pq.
What are the differences between SPF-p.q and RBF-pq?
One of the factors is a between-block effect (or betweensubje

Randomized Block Factorial Design (RBF-pq)
(Kirk, pp.456-461)
CR-p could be improved by RB-p.
Similarly, CRF-pq can be improved by RBF-pq by blocking
effect.
Assume n blocks
n homogeneous groups of people, or
n people with p x q repeated measures
(mod

Detailed Analyses of Interaction Effect
When the interaction effect is significant, we might be
interested in what exactly causes the interaction effect.
There are several different ways to look at the interaction
effect. For example:
1. Simple Main Effec

Intraclass Correlation
It is a effect size for the block effect.
(computational formula)
FBL 1
I Y |BL A =
( p 1) + FBL
(e.g.)
17.4 1
I Y |BL A =
= 0.89
(2 1) + 17.4
(This is an extremely high intraclass correlation)
It indicates the amount of varianc

Critique of one-way ANOVA
Model: Yij = + j + i ( j )
explained
(structure)
explained
(structure)
unexplained
(error)
unexplained
(error)
Total
variation
Structure of the model typically does not account for much
of the total variation in the outcome vari

Dunn-Sidak Test - simply Sidak in SPSS
A modification of Bonferroni procedure.
Slightly more liberal than Bonferroni narrower
confidence intervals, and more powerful.
In Bonferroni the modified -level is:
/2
C
(e.g.)
0.05 / 2
= 0.0125
2
So,
*
t * = 0.0

Effect Size Indices
Omega Squared (2)
2
2
2 +
2 =
Conceptually, it is a proportion of variance accounted for by
group differences.
We estimate 2 by
p
2
=
j =1
and use
p
2
j
=
( p 1)( MSBG MSWG )
np
2 = MSWG .
Also, a formula for estimating 2 is writte

Multiple Comparisons
Contrasts among means
Notation:
i . ith contrast among population means.
i . The estimate of i.
Example:
For 3 groups,
1 = Y1 Y2
Y1 + Y2
Y3
2
Y +Y
5 = 1 3 Y2
2
Y +Y
6 = 2 3 Y1
2
4 =
2 = Y1 Y3
3 = Y2 Y3
(4.1-1)
Coefficients for

Experimental Design as a Research Strategy
" … . A plan for assigning subjects to experimental conditions and
the statistical analysis associated with the plan."
(Kirk, 1995, p.1)
• What makes experimental research distinctive from other
research strategi