Keith
Markus' Homepage:
https://jjcweb.jjay.cuny.edu/kmarkus/
PSYC U86001.01
Syllabus
PSYC U86001.01
Structural Equation Modeling
Professor Keith A. Markus
Fall 2022
CRN: 48759
Crosslisted as: EPSY U83400.01 (52555)
Time: Wednesday 6:308:30 PM
Room: L2.72.05NB, John Jay College of Criminal Justice,
524 W59 Street, http://www.jjay.cuny.edu (Room 72 is a suite
on the north side of floor L2, look for "CLSS", turn left after
entering past the main desk. The room is easiest to find from
the 59th Street or 11th Avenue entrances.)
Office Hours:
by appointment
(Most questions can be more effectively and efficiently answered
using email or Blackboard Discussion forums.)
Contact Information:
Dr. Keith A. Markus
kmarkus@aol.com
2122378784 (If I do not answer, it is best to send email than
leave voice mail.)
Room 10.63.05 NB, 524 W59 Street. (Use elevator bank near 11th
Avenue)
Psychology Department, John Jay College
Course Description:
The course will provide a general introduction to the use of
structural equation modeling in empirical research. The course
will pay special attention to criminal justice applications,
although it will provide an appropriate introduction for
applications in any social or behavioral science. The course
will cover path analysis, confirmatory factor analysis, and
structural equation models with latent variables, including some
useful special cases. The coverage will include appropriate
research design, model specification, parameter estimation,
assessment of model fit, and model interpretation. The
treatment of these topics will emphasize practical
application. The course will also introduce the use of at
least one software package.
Course Objectives:
1. Provide a survey of introductory topics in SEM.
2. Provide a foundation for advanced topics in SEM.
3. Provide experience with conducting SEM analyses.
4. Provide cognitive inoculation against the most common errors in
applying SEM.
Diversity and Inclusion:
This course is intended to provided a supportive and welcoming
learning environment for all students. I want every student
registered for the course to get the most out of the course.
Structural equation modeling offers powerful statistical tools for
detecting biases and potential adverse impact. Although this
course is an introductory survey and cannot get to deep into any one
topic, I hope that it will provide a useful foundation and stepping
off point for anyone interested in exploring and applying such
methods. Like any statistics course, this course builds on
statistical fundamentals that have some ugly episodes in their
history. It is my hope and belief that such elements have been
thoroughly purged from modern statistical theory. Nothing in
this course assumes or implies them. I cannot promise that I
will always get everything right but I will do my very best to make
every student feel welcomed and supported in this course. If
you have any issues or concerns in this regard, please do not
hesitate to reach out to me.
Text Book:
Kline, R. B. (2016). Principles
and
practice
of
structural
equation
modeling
(4rd ed.). New York: Guilford. (The publisher's
page links to electronic supplements to accompany the book.)
Handouts: Additional handouts will be posted on Blackboard.
Required Software:
R with the lavaan
package and simsem package
installed: R is a powerful opensource free statistics environment
but requires some adjustment for those accustomed to point and click
statistical environments. (Many students like to use third party
GUIs like R Studio or R Commander.) The two packages do not
come with the base installation and must be added after you install
R. It will not be necessary to master R in order to use the packages
for the class. See the package web pages for additional
documentation beyond what is provided in R help files. If you are
completely new to R, you may want to follow along with the sample
session provided at the end of An
Introduction to R just to get used to the R environment. This
document offers a useful introduction to R, although it covers a
great deal of material not needed for this course. R has a
builtin GUI for Windows and Mac but not for Linux.
Installation:
1. Point your web browser to the Comprehensive R Archive
Network (CRAN).
2. From the sidebar menu on the left, near the top, click Mirrors
and select something geographically close (e.g., Pennsylvania).
The same page will reload from a closer server.
3. Select Windows (if that is your operating system), if you use
an Apple computer, your version of R differs somewhat and I am not
familiar with it but all the examples should run.
4. Click base. Then download and run the newest version
installation file (currently R3.3.1win32.exe). Further
installation instructions are provided on the CRAN web page.
5. Once installation is compete, start R. You will see a window
with a '>' prompt. At the prompt you may type the following
command to test the installation.
> demo(graphics)
You will be prompted to hit Enter several times as you move
through the demo. A series of graphs should appear in a separate
window inside the R window if R has been installed
correctly. The remaining instructions are specific to the
native R GUI but the procedure is very similar in RStudio.
You can also install packages from the console using the
intall.packages() function which has a help file and several
related functions. However, I do not recommend this for
beginners.
6. On the Packages menu in R, select Install Packages. You will
be prompted with a list of mirror sites that opens in a separate
window. Again, pick something close (e.g., USA PA or USA PA2).
7. Momentarily, you will be prompted with a list of packages in a
window similar to the mirror site window that you just used. Click
the package name (e.g., lavaan).
8. After some brief chugging, you should have a message in your
main R window indicating that the package installed correctly.
9. You can test the installation by typing the following command
at the R prompt.
>library(lavaan)
>?lavaan
This should open a new window outside of the main R window with a
help file on the lavaan() function.
10. Return to the R console window where you type commands. At the
prompt, enter the following command. When prompted, choose not to
save the workspace image. This will close R.
>q()
Repeat steps 69 for the simsem package. Use library(simsem) and
?sim to test the installation.
Why not, one might wonder, use popular commercially available SEM
programs such as LISREL, EQS, AMOS or Mplus? Because they are very
expensive and free student versions typically only have the
ability to run very limited models. Once you learn to use the free
packages above, you will be in a better position to evaluate which
SEM software you might want to purchase. You will also have a
sufficient foundation to make the adjustment to alternative SEM
software.
Blackboard Access: Access
to Blackboard is an essential part of this course. Course materials
will be distributed through Blackboard and I will use Blackboard to
send you email. If you have any difficulty accessing the Graduate
Center Blackboard system, please resolve those difficulties as soon
as possible.
Course Meetings: The room used for this course is a
computer lab. Lecture material will be posted on
Blackboard. We will use class time primarily for exercises
designed to reinforce the material from the reading.
Selfintroduction Assignment: Use the "Introduce Yourself"
discussion forum on Blackboard to post a brief introduction.
Indicate what program you are in, why you are interested in SEM, and
your research interests. (This will be graded as complete,
100%, as long as you submit it.)
Self Introduction:
Post a brief message to the corresponding discussion board on
Blackboard introducing yourself to the class. Content might
include what program you are in, what your research interests are,
and what you hope to get out of the course.
Course Project
As a course project you will conduct a small simulation study.
The project is broken into three pieces.
Research Design Poster
Midway through the course we will start to study model specification
for various classes of models. At that point, you should have
what you need to decide on a research design. Your design
should incorporate at least 4 conditions and at least two dependent
variables (but they can be two measures of the same conceptual
outcome). The four conditions can result form a single
independent variable or a factorial crossing of two or more. I
strongly recommend reading the simulation tutorial before completing
this assignment. See the Software and
Downloads part of my web page for an overview of some common
research designs and sample code for each. It is very
important that you clearly distinguish the research design that you
are simulating from the research design of the simulation itself
(both in your thinking and in your writing). Please also read
the document "Writing About Simulation Studies" before completing
this assignment.
Do not use the same mediation model from the tutorial for your
simulation. Use your own model.
Create a conference poster presenting your research design.
See the APA Handout offering guidance on poster design. The
poster should be 4 feet tall and 6 feet wide. Presentation
software like PowerPoint or Impress offer a good options for
creating posters. After modifying the default page size, you
can create the poster on a single sheet. Whatever method you
use to create the poster, please turn it in as a pdf (portable
document format) file to ensure that I can see it the way you want
it to look.
Organize your poster as a research proposal with an introduction
section and a method section. The introduction should cite
some related literature but need not be comprehensive, a few
representative sources will do. Divide the method section into
subsections. Clearly label each of these with a
subheading. Include any references on the poster. Your
research design should specify at least the following information
about the simulation study. Note that sections (a) and (b)
refer to variables in your simulation study design, not variables in
your model. Variables in your model belong in section (d).
(a) Independent variable(s), levels or range of values if continuous
and how it will be manipulated.
(b) Dependent variable(s) and how it will be computed. Items
(a) and (b) refer to variables in your simulation design, not
variables in your model.
(c) Sample size (this refers to the number of replications, not the
number of observations simulated for each replication).
(d) Model specification (for some studies, this may vary by
condition). It is okay to include a path diagram but do not
rely on that exclusively. Give both equations and the lavaan
model syntax.
(e) Hypotheses or research questions.
Some common dependent variables include the following.
(a) Type I error coverage: Do 95% of the parameter estimates fall
inside the 95% confidence interval? Would a null hypothesis be
rejected at the nominal rate when true (e.g., 5%)?
(b) Bias: What is the mean difference between the estimates and the
population value?
(c) Relative bias: mean((Estimate  Value) / Value), this
gives the amount of bias relative to the size of the population
parameter value.
(d) Mean goodness of fit statistics.
(e) Statistical power (rejection rate for false hypothesis).
(f) Proportion of successful convergence on a proper solution.
Test Run Poster
We will use the simsem package in R to run the simulations.
The simsem package permits model specification using lavaan syntax
and model estimation using the lavaan package. Also see the
simsem online documentation which contains many examples.
(Focus on the examples using lavaan syntax to specify models, simsem
allows other methods too.)
Run one replication of one condition in your design. Provide
(a) the code for your test run, (b) the output from the summary()
function, and (c) the output from the summaryTime() function.
Review the output and correct any apparent problems before copying
into your poster. Provide a brief verbal description of the
condition that you ran. Interpret the output in terms of the
success with which you implemented the condition. Revise the
material from the research design to reflect any changes that you
have made in the design. You may also wish to shorten the
presentation of the material from the research design to make room
for the new material. Note: If some conditions in your design
are more complicated than others, I would suggest running the most
complicated condition. If that runs, then the simpler
conditions should also run, but the reverse is not true. (You
are welcome and encouraged to test run all of your conditions but
only include one condition in the test run assignment.)
Research Report Poster
Make any needed further modifications to your introduction and
method section. Add a results section that presents the
results of your simulation study and a discussion section that
connects the results to the literature discussed in the
introduction.
Grading: Your course grade = .25 (Research Design) +
.25 (Test Run) + .40 (Research Report) + .10(Self
Introduction). Letter grades will be assigned as indicated
below.
Letter Grade

Percent Grade

A

92100

A

8491

B+

7683

B

6875

B

6067

C+

5259

C

4451

C

3643

F

035

Special Needs:
To request accommodations please contact the Office of the Vice
President for Student Affairs (Room 7301 Graduate Center; (212)
8177400). Information about accommodations can be found in the
Graduate Center Student Handbook 0506, pp. 5152).
Academic Honesty:
The Graduate Center of The City University of New York is
committed to the highest standards of academic honesty. Acts of
academic dishonesty include—but are not limited to—plagiarism, (in
drafts, outlines, and examinations, as well as final papers),
cheating, bribery, academic fraud, sabotage of research materials,
the sale of academic papers, and the falsification of records. An
individual who engages in these or related activities or who
knowingly aids another who engages in them is acting in an
academically dishonest manner and will be subject to disciplinary
action in accordance with the bylaws and procedures of The
Graduate Center and the Board of Trustees of The City University
of New York.
Each member of the academic community is expected to give full,
fair, and formal credit to any and all sources that have
contributed to the formulation of ideas, methods, interpretations,
and findings. The absence of such formal credit is an affirmation
representing that the work is fully the writer’s. The term
“sources” includes, but is not limited to, published or
unpublished materials, lectures and lecture notes, computer
programs, mathematical and other symbolic formulations, course
papers, examinations, theses, dissertations, and comments offered
in class or informal discussions, and includes electronic media.
The representation that such work of another person is the
writer’s own is plagiarism.
Care must be taken to document the source of any ideas or
arguments. If the actual words of a source are used, they must
appear within quotation marks. In cases that are unclear, the
writer must take due care to avoid plagiarism.
The source should be cited whenever:
(a) a text is quoted verbatim
(b) data gathered by another are presented in diagrams or tables
(c) the results of a study done by another are used
(d) the work or intellectual effort of another is paraphrased by
the writer
Because the intent to deceive is not a
necessary element in plagiarism, careful note taking and record
keeping are essential in order to avoid unintentional plagiarism.
For additional information, please consult
“Avoiding and Detecting Plagiarism,” available in the Office of
the Vice President for Student Affairs, the Provost’s Office, or
at
http://web.gc.cuny.edu/provost/pdf/AvoidingPlagiarism.pdf.
(From The Graduate Center Student Handbook 0506, pp. 3637)
Schedule
Date

Reading Assignments Due

Project Assignments Due

Week1: W 8/31

Introduction &
Chapter 1: Coming of Age
Handout: SEM glossary.
Overview of SEM and course. Theories and
models. Installing R.


Week 2: W 9/7

Chapter 2: Regression Fundamentals &
Chapter 3: Significance Testing and Bootstrapping
Introduction to using R.

Self Introduction

Week 3: W 9/14

Chapter 4: Data Preparation and Psychometrics
Review.


Week 4: W 9/21

Chapter 5: Computer Tools
&
Chapter 6: Specification of Observed Variable (Path)
Models.
Handout: SEM Matrix Algebra Basics.


Week 5: W 9/28

Chapter 7: Identification of
ObservedVariable (Path) Models. 

Week 6: W 10/12
(no classes 10/5)

Chapter 8: Graph Theory and
the Structural Causal Model 

Week 7: W 10/19

Chapter
9: Specification and Identification of
Confirmatory Factor Analysis Models.


Week 8: W 10/26

Chapter 10:
Specification and Identification of Structural Regression
Models. 

Week 9: W 11/2

Chapter 11: Estimation
and Local Fit Testing

Research Design

Week 10: W 11/9

Chapter 12: Global Fit
Testing


Week 11: W 11/16

Chapter 13: Analysis of
Confirmatory Factor Analysis Models. 

Week 12: W 11/23

Chapter 14: Analysis of
Structural Regression Models. 
Test Run

Week 13: W 11/30

Chapter 15: Mean Structures
and Latent Growth Models


Week 14: W 12/7

Chapter 16: MultipleSample
Analysis and Measurement Invariance.
Chapter 18: Best Practices in Structural Equation Modeling


Finals
Week: W 12/21


Research
Report

Created January 28, 2008
Migrated 17 August 2011
Updated 25 August 2022
This page was created using
SeaMonkey
v.2.53 and is best viewed using a Mozilla web browser.