Background
While on the marketing faculty at Arizona
State University, I had the privilege of teaching a doctoral
seminar in multivariate statistics. That I was teaching
the course probably surprised me more than anyone. Why?
Because I've never considered myself a 'quant jock.' For
reasons probably due to genetic ancestry, I find it difficult
to get excited about the inner workings of optimization
algorithms or exploring the sensitivity of MANOVA to violations
in the independence assumption, for example. Rather, my
interest in the various multivariate tools arises from their
usefulness as a means for examining phenomena that do interest
me. Now, if an analysis tool allows me to identify the direct
and indirect effects of mass media on identity vs. the direct
and indirect effects of possessions on identity, you've
got my attention! (can you identify an analysis tool that
can do just this?).
As a number of folks have asked for the
seminar's syllabus and/or reading lists, I've published
them here for easy access by all (well, at least all with
ready web access).
If you have any questions, please drop
me a note.
A Caveat
I am not a quant jock, nor do I play one
on TV. My strength with multivariate statistics is in my
ability to explain multivariate analysis tools in a way
that makes sense to the statistics-challenged. I emphasize
the issues scholars face in trying to divine whether the
data support their a priori hypotheses.
An Emphasis
In my experience, few multivariate statistics
courses emphasize, or even address, the practical issues
associated with doing the tasks fundamental to the success
of any research project:
- designing a study with full consideration
given to the analysis methods used to test the hypotheses,
- applying the analysis tools
to evaluate measurement quality, and, ultimately,
- applying the analysis tools to
your data and interpreting the results in light of your
hypotheses.
I designed this seminar to address head-on
those weaknesses.
Assumptions
This semester-long seminar was designed
originally for marketing doctoral students who had completed
all other program course work and were rapidly approaching
their comprehensive exams and the task of developing their
dissertation proposal. Students were assumed to bring with
them familiarity with the research process, experimental
design, ANOVA, and multiple regression. The content of the
course proved popular with students in not only marketing,
but also management, operations research, and political
science.
Seminar Overview
The seminar flows like this:
We begin with an overview of the research
process and issues unique to the dissertation process.
Measure Quality Assessment then receives
extensive coverage. We begin exploring measure reliability
and validity with simple bi-variate correlations, factor
analysis, and ultimately confirmatory factor analysis.
Testing Construct Relations is then covered
extensively. We first build on our knowledge gained via
confirmatory factor analysis to explore hypothesis testing
using the full covariance structure analysis framework.
We then tour the more conventional multivariate analysis
tools (e.g., MANOVA, multivariate regression, etc. with
an emphasis on how to use them to test theoretically driven
construct relations.
|