Jörg Blasius
Correspondence Analysis
Jörg Blasius is Assistant Professor at the Zentralarchiv für Empirische Sozialforschung, University of Cologne. His research interests are exploratory data analysis, methods of empirical social research, and the sociology of lifestyles.
This course introduces the data reduction technique of correspondence analysis (CA). Ca is a multivariate method for exploring cross-tabular data by converting such tables into graphical displays, called “maps”, and related numerical statistics. The course starts with a detailed introduction to the geometric background of the method. Using several applications from the social sciences, the symmetric and asymmetric presentation of maps as well as the numerical results will be examined in detail. Besides simple and multiple correspondence analysis, the course will give an overview of joint correspondence analysis and related scaling techniques, such as principal components analysis and multidimensional scaling. Furthermore, there will be discussion of the analysis of various types of data via CA, for instance time dependent data such as panel and trend data and multi-response data. Finally, the course compares log-linear analysis as a model-driven approach to analyzing categorical data with an exploratory type of data reduction such as CA.
Participants are invited to bring their own data.
The Software used are SimCA SimCA (a stand-alone program written by Michael Greenacre)
and SPSS.
Bibliography
Basic text/overview
Greenacre, Michael 1993. Correspondence Analysis in Practice. London: Academic Press.
Greenacre, Michael and Blasius, Jörg 1994 (eds.) Correspondence Analysis in the Social Sciences. Recent Developments and Applications. London: Academic Press.
Blasius, Jörg and Greenacre, Michael 1998 (eds.) Visualization of Categorical Data. London: Academic Press.
Remedial Reading
Kim, J. O. and Mueller, C. 1979. Factor Analysis. Sage. QASS, 14. (or any other introduction to principal components analysis)
Prerequisites
A working understanding of descriptive statistics is sufficient, but it would be helpful if participants have some understanding of linear algebra, multidimensional scaling and principal components analysis. An elementary knowledge of inferential statistics is recommended, especially the chi-square-test of independence for a contingency table.