Correspondence Analysis

Short Bio Philippe Blanchard, PhD in political science, is an Assistant professor in political science at the University of Warwick, UK. He works on green politics, political communication, and methods for social and political sciences: multivariate statistics, longitudinal methods, interviewing and content analysis. He is presently conducting research about gendered careers in trade unions, transnational economic elites, and big data for political science. He has taught methods and techniques for social and political sciences in France, Switzerland, Austria, Denmark and the USA. http://www2.warwick.ac.uk/fac/soc/pais Prerequisite knowledge Participants should have the basic skills for social sciences statistics, that is, some knowledge of: - Windows or Linux. - Excel or equivalent spreadsheet software. - Questionnaires: cleaning, coding, datasets management. - Basic mathematics: geometric space, vectors - Descriptive statistics: tabulations, summary statistics, chi2, correlation - A standard statistical package: SPSS, SAS, Stata, R or other. Participants should also be willing to learn another approach to social and political statistics. Correspondence analysis brings another angle to social reality, yet fully complementary to more usual methods, including qualitative-comparative analysis (QCA), regression models, cluster analysis or content analysis. In any case, if you wonder if the course fits your research project or if you match the prerequisites, or if you need reading advices, do not hesitate to ask the instructor: p.blanchard@warwick.ac.uk. Short course outline Correspondence analysis is the core method of a tradition of multivariate descriptive statistics also known as (French/geometric/structured) data analysis. It is unrivalled in the study of political values and attitudes, sociological and ideological profiles of electorates, and organisational memberships. It enables in-depth exploration of complex datasets, combined with geometric representation. It relies on a come-and-go process between data at aggregated and individual levels. Correspondence analysis is particularly adapted to the treatment of non-continuous data and multiple choice questionnaires (about tastes, belongings, experiences, etc.). Although complementary to regression analysis, it is more at ease with nominal data than regression. More resistant to complex dependence between variables, it authorizes more flexible causal designs, based on thorough empirical description. The course will rely on basic statistical skills (see prerequisites), geometric intuition and knowledge of the context of the study. It is devised for beginners, but students already using correspondence analysis and in search of a more rigorous practice or advanced techniques may also participate.