Roger Millsap's Home Page
This website contains links to various pages related to my
teaching, research, publications, and personal interests.
I am a faculty member in the
Department of Psychology
at Arizona State University in Tempe, Arizona. My research focus is in
quantitative psychology, which means that I develop and/or apply
statistical and mathematical methods to help resolve research
questions in psychology. I am currently teaching coursework in
psychological measurement and statistical methods at the undergraduate and
doctoral levels as an active member of the
Quantitative Psychology Doctoral Program in the Department.
I am also working as the Co-Director of
the Methodology Core group at the
Prevention Research Center
on campus. This Center conducts research on programs to help
children and families cope with stressful experiences, with a focus
on parental divorce, poverty, bereavement, and parental job loss.
For a general description of the field of quantitative psychology, a list
of doctoral programs, how to get into the field, and so forth, go
If you want to find out more about what topics or questions are addressed in
this field, Albert Maydeu-Olivares and I edited a
that surveys current methods in the field of quantitative psychology.
My publications are listed
here. I now serve as the Editor of
Psychometrika, the journal of
the Psychometric Society. Psychometrika is a journal that
publishes papers on all aspects of quantitative psychology.
I am on
the Editorial Boards of three other journals,
Multivariate Behavioral Research,
Applied Psychological Measurement,
Structural Equation Modeling.
I am active in several professional organizations:
My central research interest at present concerns methods for detecting bias in
psychological measures. The bias of interest concerns systematic errors of
measurement that are associated with group membership, with groups usually
defined demographically (although other definitions of "group", such as
randomized groups in experiments, are potentially interesting as well).
Technically, we say that a measure is "biased" if systematic group differences
in observed scores exist among individuals who are identical on the attribute(s)
being measured. This latter requirement is the difficult part, as there is
ordinarily no sure way to completely "match" individuals on attributes that cannot be
directly measured. Nearly all psychological attributes fall in this
category; we cannot directly measure emotions, attitudes, or abilities.
The challenge of this research area is to develop methods for studying bias
in the absence of simple, direct measures on which to match individuals from
Related to the above, I have an ongoing interest in statistical methods
associated with latent variable models , especially factor analysis
and item response theory. Simply put, latent variable models attempt to
model observed states or conditions as outcomes of unobserved processes
or conditions. A latent variable model imposes some structure in the
relationship between the observed and unobserved variables, but the
structure is not generally fully amenable to empirical tests. Factor analysis
is one of the older models (see
FA100 for a recent conference on
100 years of factor analysis); item response theory is a bit younger, but
hardly new. Aside from applications of latent variable models to the
measurement bias problem, I have done work in longitudinal models,
factor models for multitrait-multimethod data(MTMM), and identification problems
in latent variable models.
For a list of resources related to the above research interests, see the following:
- For measurement invariance and my book "Statistical Approaches to Measurement Invariance"
- For structural equation modeling, go here...
- For psychometrics, go here...
I am currently teaching several courses on a rotating basis:
- Psych 534 Psychometric Methods is a doctoral course in the theory and practice of psychological
- Psych 591 Advanced Psychometrics is a course in item response theory (IRT), but also dwells on
the use of confirmatory factor analysis to test measurement hypotheses. PDF.
- Psych 533 Structural Equation Modeling is a course in structural equation modeling (SEM).
The course assumes no prior exposure to this topic, but does require a previous course in multivariate statistics.
Last Updated April 17, 2011
- Buddhism? Go
- Books, Literature, Bookstores, Search Engines? Go
Web Page by Roger E. Millsap
Send Roger an email here.