Multivariate Statistics 1920
Multivariate Statistics 1920
Political Science MA
Multivariate Statistics
Winter 2020
Instructor office hours: 14:30-15:30 Mondays and Thursdays Please send an e-mail if you
wish to come
Learning goals
The course develops, on the one hand, a theoretical understanding of why and how
multivariate statistical methods are applied in political science and, on the other hand, skills,
including the application of statistical packages, to carry out such analyses and the
interpretation of their results. The course prepares the students to critical application of
multivariate statistical methods and requires oral and written presentation of the results.
Precise interpretation of the results of the analyses will be emphasized, throughout. The
presentation of the statistical methods will concentrate on the research question they may
answer, and variations depending on characteristics of the data, including the level of
measurement, will be discussed as a second level of differentiation.
Univariate regression
Multivariate regression
Logistic regression
Log-linear modeling
Factor analysis
Cluster analysis
1.09. Introduction: the role of multivariate statistical analysis, questions which may be
answered with statistical analysis, data collection, experiments and observational studies,
levels of measurement
1.13. Review of univariate regression analysis, the statistical question, level of measurement
of the variables, model fit, parameter estimates
1.20. Model fit, test of effect, the role of normality, nonparametric variants
1.30. Stepwise model selection, sources of misfit, ANOVA around the predicted values
2.10. Log-linear models for categorical data: conditional and higher order odds ratios,
association and simplicity, log-linear model interpretation, regression type models
2.24. Reduction of dimensionality for numerical variables: principal component and factor
analysis
3.16. Discussion of the final projects, and time reserve for material not covered
Texts
www.rcommander.com
http://cran.r-project.org/doc/contrib/Faraway-PRA.pdf
http://little-book-of-r-for-multivariate-analysis.readthedocs.org/en/latest/
http://cran.r-project.org/doc/contrib/Karp-Rcommander-intro.pdf
Reading and Understanding Multivariate Statistics / Laurence G. Grimm and Paul R. Yarnold
Reading and Understanding More Multivariate Statistics / L. G. Grimm and P. R. Yarn
Students are not expected to read these books. Rather, the books may be consulted on
particular topics.
Software
The class is software neutral. This means that students are welcome to run the statistical
analyses discussed on any software of their choice. The instructor will use R for classroom
presentations, with the graphical user interface R Commander. Support will be provided to
students to implement these free pieces of software and making their own data readable by R.
Class procedures
The class relies on active student participation to make sure each student gains a clear
theoretical knowledge and hands-on experience. The in-class presentations by the instructor
will cover theoretical aspects illustrated by simple applications. After each class, a very brief
summary of the material covered will be posted on the e-learning site. Students will write a
midterm test, submit assignments and work on a project. Students are required to form groups
to help each other in completing their projects. Each group will work on the analysis of a data
set of their choice. While joint work is encouraged, each student will have to have an
identifiable contribution to it. In particular, each student will have to choose an analytical
problem, contribute to a work-in-progress presentation and give a final presentation in class
and submit a final written research report.
Students who audit the class will have to successfully write the midterm test