Over the past three decades, sociologists and statisticians have made
dramatic strides in developing a growing array of powerful methods
for the analysis of data arrayed in the form of contingency tables
and dependent variables measured at the ordinal or nominal-level.
Categorical data of this sort are extremely common in contemporary
sociology and related disciplines, and in this course we examine in
detail a number of models and techniques applicable to the analysis
of such data. Our focus will favor depth over breath so as to develop
a detailed understanding of several powerful types of models that are
particularly promising for research applications. We will emphasize
data analysis throughout the seminar, including issues of
interpretation, presentation of results, and
using statistical models to connect theoretical questions and
empirical data.

The first third of the course surveys log-linear models used to
analyze association, structure, and symmetry in contingency tables.
These models figure centrally in the analysis of stratification,
mobility, and family behaviors. The second part of the course surveys
models of latent structure. In addition to stratification
applications, these models are used by opinion researchers and social
psychologists and figure heavily in debates over attitude measurement
and change. The third section of the course reviews the binary logit
model, considering its multinomial extensions and some of the
statistical problems that arise in modeling choice behaviors. We also
investigate some useful techniques for
categorical data models in which time of data collection is itself a
covariate.

Throughout the course, I will provide examples and exercises in which
we analyze data from a number of classic and contemporary surveys.
Although these analysis are generally motivated by substantive
research questions (some of which represents ongoing scholarly
debates), students in S651 are also encouraged to bring with them or
seek out a concrete research question and source of appropriate data
with which to develop new
analysis.