Course Description: The course offers a general
introduction to psychometric methods primarily emphasizing
classical test theory, test construction and validation, and test
use. The emphasis lies with developing a firm understanding
of basic psychometric concepts. This course lays a
foundation for more advanced courses in specific topics introduced
here. The course understands psychometrics and testing as applying
broadly, not just to paper and pencil tests but also to
performance assessments, behavioral observations, measured
variables in experiments and quasi-experiments, surveys, and other
forms of behavioral data collection. However, much of the material
will emphasize measurement involving multiple indicators of a
common construct.
Course Objectives: The
course assumes a foundation in basic statistics and a healthy
curiosity but little more. The more you put into the course, the
more you will get out of the course. The course design reflects
the following objectives.
1. Students will gain a basic understanding of the foundations of
test theory that will prepare them to pursue more advanced topics
(e.g., item response theory, structural equation modeling).
2. Students will gain the background and confidence to critically
read technical manuals and other documentation in conjunction with
use of published tests.
3. Students will gain facility with conceptual tools for thinking
through issues of validity and reliability as applied to all
measures from dependent variables in experiments to large scale
testing programs.
4. Students will gain a level of comfort with algebraic
representations of test scores and the use of these to think
through applied problems related to test use and interpretation.
5. Students will gain an increased sensitivity to the fallibility
of educational and psychological tests and the limits to their use
and interpretation.
6. Students will gain exposure to the use of statistical software
for conducting psychometric analyses and some experience with such
analyses.
Text Book:
Draft chapters by Professor Deborah
Bandalos will be provided for reading. These are
provided for your personal use in the course only. Do not
share them with anyone outside the class. This condition was
stipulated by the publisher in order to make it possible for us to
use the draft chapters this term. Please return the kindness
of the author by not sharing the draft chapters.
I will illustrate psychometric concepts using a variety of
software packages. Familiarity with the software is not a
course requirement. However, learning psychometrics simply by
reading about it is akin to learning to swim, ski, or play a
musical instrument simply by reading about it. Actual practice is
a much more effective method. Whether you use a simple calculator,
a spreadsheet, or advanced statistical software, it is a good
habit to play around with the material by constructing concrete
examples and taking a try-and-see attitude toward the material. If
something seems puzzling, make up an example and try it out. If
something seems counter-intuitive to you, try to construct a
counterexample. The more concrete you make psychometrics, the more
comfortable you will feel with the material, the better you will
understand it, and the more skills you will develop that you can
apply outside of the class. None of the this is required for the
course, but it will make it more fun, more interesting, and more
valuable at a practical level.
Homework: Turn in your chapter
review electronically before the start of class each week.
Chapter reviews do not apply the chapters from the
Standards. For each chapter not from the Standards, turn in
a text or rich text (rtf) document with the following three
questions followed by your answers for the chapter in
question. Copy and paste the questions into our document and
type an 'X' over the number for your rating.
1. What one part of the chapter would you most like to see
improved and why?
2. What one part of the chapter would you most like to see kept as
it is and why?
3. Rate the chapter on the following scale and briefly explain
your rating, citing examples if applicable.
(1) No new information for me
(2)
(3) Too easy
(4)
(5) Just right
(6)
(7) Too hard
(8)
(9) Completely incomprehensible
Course
Project:
Week 3: Project Topic
Post your choice of test population and 2 constructs to
Blackboard. Include all three pieces of information in the subject
line. Do not choose something that has already been taken. State
your choice with sufficient specificity that others can still
choose distinct but related options. The test population
constitutes the population of test takers for whom you intend the
test. The two constructs should be unidimensional quantities
across which you can expect variability in the test population.
Week 7: Project Part I
Begin a proposal for the validation of a test of your construction
following the format below. Use the following lettered headings to
clearly mark each section of your proposal. Double space the
proposal and use APA format and style. However, printing on both
sides of the page is fine. Turn in a paper copy at the start
of class.
A. Title page including your name and affiliation.
B. Abstract (180 words max). (Only include Part I material.)
C. Purpose of the test (250 words max). Describe the intended use
of the test. Describe the intended users and the intended test
taking population. Explain what the test would contribute over and
above existing tests. Describe the theoretical rationale behind
the test.
D. Test Blueprint (500 words max).
1. Define the constructs to be assessed by the test. Your
test should include at least two constructs and at least six items
per construct. Describe the relationship(s) between the
constructs, conceptually and statistically.
2. Specify the format of the items and response options.
3. Specify the content of the items. If a scale on your
test includes more than one kind of item, specify the number of
items of each type.
4. Specify the acceptable range of item statistics (mean or
proportion correct, standard deviation) for each item and test
statistics (mean, standard deviation, reliability) for each
subscore.
Week 11: Project Part II
Turn in a paper copy at the start of class.
Add the following to your proposal.
A. Expand the abstract to include parts E and F.
E. Draft test. Provide a draft version of the test including
instructions and a full set of items that conform to parts 1-3 of
the test blueprint.
F. Proposed validation plan (1000 words max).
1. Outline your validity argument for each scale by stating
the desired interpretation and listing supporting assumptions (250
words max). Mark the assumptions that you consider essential
to the interpretation of the test scores and then rank them in
terms of degree of support. Identify the five least well
supported essential assumptions.
2. For each of the five assumptions identified in section
F1, describe a validity study to test that assumption (750 words
max). Keep each study individually focused on one source of
evidence, but try to design a set of studies that collectively
provide a range of different types of validity evidence.
Clearly identify the subsection for each study using subheadings.
Week 15: Project Part III
Convert your paper to portable document format (PDF), and turn in
both a hard copy and a PDF file. (If you cannot save directly to
PDF, download a free PDF print driver such as Cute PDF.)
Add the following to your proposal.
A. Expand the abstract to cover the entire proposal.
G. Factor model (250 words max).
1. Present a 2-factor model appropriate to our test
(present this in theory, this task does require conducting a
factor analysis of data). Choose plausible values for the item
parameters (loadings, unique variances, intercepts, factor
correlation[s]). Interpret the model in terms of which items
assess which constructs and how well.
2. Use the factor model to simulate data as if it came from
your test (use any software you like, spreadsheet or statistical
software). Simulate at least 100 cases.
3. If necessary, tweak your factor model values until your
items satisfy part 4 of the test blueprint.
H. Pilot study (500 words max). Report this as if you had
completed an empirical pilot study, but use the simulated data
from part G. (Because your subscales measure different things, you
will want to analyze them separately. However, if a total score
has meaning for your test, it may also give you something
interesting to write about if you analyze the scale as a whole as
well.)
1. Conduct an item analysis of the data set. Report the
item statistics (means, standard deviations, item total
correlations and regression R-square values). Describe how
differences between item statistics relate to differences between
item parameters in the factor model.
2. Report the scale statistics (means, standard
deviations). Compare and contrast the scale statistics for each
scale.
3. Report both alpha and lambda 2 reliability estimates.
Describe the alphas-if-item-deleted.
4. Relate the results from parts H1 to H3 to the test
blueprint. Provide an overall evaluation of the functioning of the
draft test based on these results.
J. Appendices: Include the spreadsheet or syntax used for your
simulation as Appendix A, and the item anlysis output as Appendix
B.
Proposals will be graded using the following rubric.
Completeness (50% of grade, 13 points total)
A&B = 1 point
C = 1 point
D = 2 points. Each of four sections = .5 points.
E = 1 point.
F = 3 points.
G = 1 point.
H = 2 points. Each of four sections = .5 points.
J = 2 points.
Overall quality dimensions (50% of grade, 40 points total)
Clarity of presentation (1 - 10)
Technical accuracy of reporting (1 - 10)
Depth with which issues are presented within allowed space (1 -
10)
Overall conceptualization and design of proposed test and test
development (1 - 10)
1-5 = unsatisfactory.
6 = minimally satisfactory.
7 = some significant weaknesses.
8 = generally good with a few weak points.
9 = overall very well done.
10 = outstanding effort.
Grading: The course project is
worth another 70% of your grade. The topic is worth 10% and each
of the three parts is worth 20%. That leaves 30% for the homework assignments. Letter grades will be
assigned as indicated below.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|