Adv Multivariate Data Analysis II -- SS08-CEP-935-001
Raudenbush, S. W. & Bryk, A. S. (2002, 2nd Edition). Hierarchical linear models in social and behavioral research: Applications and data-analysis methods. Newbury Park, CA: Sage.
Many research problems in the social sciences focus on the growth in knowledge and skills of individuals and of groups. In educational research, for example, children's growth in various natural settings such as the classroom and the school is typically the object of inquiry. However, understanding growth in organizational settings is fraught with difficulties under standard uni-level regression analyses. In fact, how to measure change and describe contextual effects on change, that is how to model nested processes, are two of the most troublesome and persistent methodological problems in the social sciences. This course is devoted to understanding these difficulties and their possible resolution using a multilevel, or hierarchical, modeling framework. The course will consider the statistical foundations of multilevel linear models, also known as hierarchical linear models (HLMs), and focuses on their application in behavioral and educational research.
Students will examine in detail a variety of multilevel or hierarchical models appropriate for a broad range of applications. The class will begin with an introduction to the hierarchical linear model (HLM), relating the HLM to the general linear model (e.g. Regression & ANOVA). The focus on the linear model will be followed with coverage of non-linear models, growth models, and alternative designs (e.g., cross-nesting—student within neighborhoods and schools). Topics discussed within the context of each multilevel model include hypothesis testing, evaluation of model fit, and computer packages that can be used to estimate the various multilevel models. Students will be exposed to selected advanced topics to be determined by the instructor.
Students should have completed CEP 933. CEP 934, a course on multivariate analysis, is a strong plus. Students who have not completed these courses must be comfortable with multiple regression, including the interpretation of output, basic formulas and conceptualization of calculations, statistical inferences, and applications to substantive issues. Although knowledge of matrix algebra is not required as a prerequisite for the course, this notation simplifies the presentation of more complex multilevel models. YOU ARE RESPONSIBLE FOR ALL YOUR COMPUTING NEEDS.
SoftwareBoth SAS and HLM software will be used in this course. SAS is available on most campus computers. A student version of HLM can be downloaded for free from Scientific Software International (http://www.ssicentral.com/hlm/downloads.html). If you would like the manual for the full version of HLM, you can order it from this website.
Grades & Grading Policy
Read the following very, very carefully:
Your grade, scored from 0 to 100, will be based on the performance of 5 homework assignments, and a final project/presentation, according to the following weights:
Students are to work in groups of three, with membership determined by the end of week one. All assignments, and also the final project, are to reflect the "best-practices" of co-operative learning. Be sure to participate as an independent contributor, rather than a "sleeping partner" (business term). If you are not contributing in a constructive manner, fellow members have the obligation to inform the instructor of your situation as early as possible.
Print and submit your homework on time. Chronically late assignments or take-home exam will diminish their collective weight to be determined at the discretion of the instructor (we really do not want that).
The final project will consist of a 6-page (text, maximum) paper addressing a HLM analysis of a substantive nature that supports your group's in-class presentation. The presentation itself will be scheduled during the final exam-day for the class. Various parts of the project will be due throughout the semester, with the paper and presentation due at the end of the course. If you do not have a dataset for your project by the end of the third week, please consult the TA and the instructor for help. Note that the presentation itself counts for 60% and the paper to be submitted weighs about 40% towards the final project grade.
Letter grades are based on the following conversion from its original score.
How to do well:
The TA will take attendance before each lecture. Punctuality is also a matter of courtesy. Send the TA a note if you know you are going to be absent, or be late.
IMPORTANT: The student, by taking the course, pledges adherence to the above said policies.