Sociology | Topics in Quantitative Sociology
S651 | 20333-20334 | James

Topic:  Hierarchical Linear Models

Many research problems of interest to social scientists posit causal
linkages at more than one level of analysis such as individuals and
groups or cities and states.  For example, the academic performance
of students is a function of both (1) the individual capacities of
students and (2) the characteristics of the schools they attend.  Are
differences in studentsí academic performance due to differences in
the qualities of schools or difference in the aptitudes of students
who attend those schools?  Questions of this kind, involving causal
processes at 2 or more levels of analysis can be properly modeled and
estimated using Hierarchical Linear Modeling (HLM) techniques.

Pervious analyses of quantitative hierarchical data sets have often
used ordinary least squares regression, which produces estimates with
poor statistical properties.  During the past 5 years, advances in
statistical estimation techniques have finally made it possible to
estimate hierarchical models properly in ways that take the
hierarchical nature of the data into account.  Given the wide variety
of applications in the social sciences, Multilevel Modeling is
perhaps the most important new statistical technique to appear in the
past 5 or 10 years.  Research papers using the technique are now
quite common in professional journals.

This is a graduate level course in applied statistics that builds
upon the ordinary least squares regression techniques.  Consequently,
students should have completed S554 (Statistics for Sociology I) or
an equivalent course in single-equation regression techniques.  If
you have not taken a course in linear regression, you should talk
with me before enrolling in the course.

Lectures will be followed immediately by a statistics lab.  We will
use the labs to gain practical experiences in the estimation and
interpretation of hierarchical linear models using the statistical
program.  Since the course will be conducted in part like a
practicum, the labs are scheduled to make sure that students have
access to the software needed to practice multilevel modeling
techniques.  Free time will be available to students to work on
problem sets.

Tom A.B. Snijders and R.J. Bosker.  1999.  Multilevel Analysis. Sage.
S.W. Raudenbush, et al. 2004.  HLM 6: Hierarchical Linear and
Nonlinear Modeling.  Chicago: Scientific Software International.

Some recommended texts will also be available in the bookstore.
These will include:
Raudenbush and Bryk.  2002.  Hierarchical Linear Models (2nd
Edition). Sage.
Kreft and de Leeuw.  1998.  Introducing Multilevel Models. Sage.