|
Lectures
|
|
Summer
Lectures
Lecture #1
1. Course Organization and class schedule
2. What is an experimental design?
Experimental Design refers to a plan for
assigning subjects to experimental conditions and the
statistical analysis associated with the plan. At
least five activities are involved in the design:
|
(a)
|
Formulation of statistical hypotheses that
are germane to the scientific hypothesis. A
statistical hypothesis is a testable
formulation of scientific hypotheses. They
are statements about either the parameter(s)
of a population or the functional form of a
population.
|
|
(b)
|
Determination of the experimental
conditions (independent variable) to be used,
the measurement (dependent variable) to be
recorded, and the extraneous conditions
(nuisance variables) that must be
controlled.
|
|
(c)
|
Specification of the number of subjects
(observational units) required and the
population from which they will be
sampled.
|
|
(d)
|
Specification of the procedure for
assigning the subjects to the experimental
conditions.
|
|
(e)
|
Determination of the statistical analysis
that will be performed.
|
3. An overview of experimental designs (Chapter 2
in Kirk)
Modern inferential statistical procedures,
especially ANOVA, are largely due to the contributions
of R. A. Fisher, J. Neyman, and E.S. Pearson who laid
the groundwork in the early 20th century. These
statisticians promoted the ideas of (a) randomization,
(b) general linear model underlying the data, and (c)
hypothesis-testing plus interval estimation as two
primary approaches to classical statistical
inference-making.
Three basic experimental designs or building
block designs: completely randomized design (or
CR-p design), randomized block design (or RB-p
design), and Latin-square design (or LS-p
design).
The completely randomized design is illustrated on
page 31 in Figure 2.2-2. The randomized block design
is illustrated on pages 34 and 35. The Latin-square
design is illustrated on page 37. Other designs are
derivatives of these three basic designs.
Factorial designs are designs in which there are
two or more main effects (or independent variables
or factors). Consequently it is possible to study the
"interaction" between these main effects. These
factorial designs are also referred to as two-way or
three-way,..., multi-way designs.
All designs share these common features:
|
(1)
|
All designs are constructed from three
basic designs;
|
|
(2)
|
There are only four kinds of variations
in ANOVA: total variation, between-group
variation, within-group variation, and
interaction variation.
|
|
(3)
|
All error terms, i.e., the denominator
of the F statistic, are formed from either
the error variation or the interaction
variation.
|
|
(4)
|
The numerator of an F statistic should
always estimate one more source of
variation than the denominator, and that
source of variation should be the one that
is under testing by 
.
|
Other terms:
Completely crossed -- A design which
includes each and every combination of all the
levels of A and B factors.
Nested design-- A design which cannot employ all
possible combinations but have levels of one factor
nested within selected levels of another factor,
such as classrooms are nested within school
buildings.
Balanced designs -- Designs in which equal
number (or proportional number) of subjects were
assigned to each treatment condition.
4. Research Strategies and the Control of Nuisance
Variables (Chapter 1 in Kirk)
Acceptable Research Hypotheses in
experimental and quasi-experimental research --
Questions that are in the form of "if A then
B".
Scientific principles in inquiry: (a) the
phenomenon must be quantifiable, (b) the phenomenon to
be studied must be passive, (c) science cannot study
things from the past, (d) scientific experiments must
be repeatable, (e) the observer or the scientist must
be at least as intelligent as the objects of his/her
scientific inquiry, (f) scientific results are neutral
but the application of the result is not, and (g)
science cannot solve moral problems.
To test a research hypothesis, several forms of
inquiry are possible (1) experiment,
(2)quasi-experiments, (3) surveys, (4) case studies,
and (5) naturalistic observations plus others (read
pages 9-16).
The purposes of these forms of inquiry are to
explore, describe or classify, establish
relationships, and to establish causality.
Four types of threats to valid conclusion based on
experimental and quasi-experimental data:
(A) statistical conclusion validity (page
17)
(B) internal validity -- the link between the
independent variable and the dependent variable
(read pages 18-19).
(C) external validity -- generalizability (page
19)
(D) construct validity of causes and effects --
operational definitions of constructs
Other threats are listed on pages 19-21.
Controlling Nuisance variables and minimizing
threats (pages 22-24).
5. Assignments:
|
(1)
|
Review Chapters 1 and 2 in Kirk.
|
|
(2)
|
Do Questions 1,2,3,4, 8, and 13 in Chapter
1 of Kirk.
Do Questions 1, 2, 4, 5, 6, 7, 9, 10, 12, 13,
and 19 in Chapter 2 of Kirk.
|
|
(3)
|
Preview Section 3.1 in Chapter 3 of
Kirk.
|
You can search for a
topic in the summer lectures by using the FIND feature
below. Replace the "..." with your search term. Do not
delete "summer and ..".
Last updated: November 24, 1998
URL:
http://www.indiana.edu/~jopeng/Y603/summer/sum1.html
Comments: jopeng@indiana.edu
Copyright
1998, The Trustees of Indiana
University
|
