Exploratory and Modelisation Techniques of Statistical Analysis for the Social Sciences 

Module de formation organisé en collaboration avec l'école doctorale "Parcours de vie".

Les 14 et 15 mai 2009, Université de Lausanne, bâtiment Génopode, salle A (quartier Sorge), 9h15-12h15, 13h30-17h

 

Journées d’étude méthodologiques sous la responsabilité de : Dominique Joye, Boris Wernli and Eric Widmer, FORS, MISC, UNIVERSITY OF GENEVA

 

Originally, back in the seventies, there was a strong separation between the statistical models elaborated in the Anglo-Saxon tradition, mostly focused on regression analysis, and the French approach “Analyse des données”.  Later on, the divide became less acute: Correspondence Analysis was included in major statistical packages, and some serious attempts were devoted to link both worlds. There is still today, however, a lively debate on the pertinence of the two  alternative models, one  causal and inferential, the other exploratory and non-inferential, for social science research, which expands to various areas of modern statistics, such as logistic regression vs loglinear models, historical event analysis versus optimal matching, structural models and latent classes versus cluster analysis, etc. This module first presents  each approach and illustrates their specificity with examples drawn from social science research. In a second part, advocates of both approaches debate together about the strenghts and weaknesses of the two approaches and about ways to make them interact.

 

Program:

  1. May 14th, 9h15-12h15: Correspondence Analysis, its strengths for social sciences (coffee break 10h30-10h50)
  2. May 14th, 13h45-17h45: Regression models,  theirs strengths for social sciences(coffee break 15h30-16h)
  3. May 15th, 9h15-12h15: Empirical applications of CA and regression models in the field of social stratification and social mobility(coffee break 10h30-10h50)
  4. May 15th, 13h45-16h: How to make both approaches meet. Round table and open discussion with the doctoral students

 

Goals of the workshop:

1)      Make doctoral students aware of the conditions of use of the two approaches

2)      Make doctoral students aware of the strengths and weaknesses of the two approaches

3)      Present doctoral students with good empirical research using both approaches

4)      Contribute to advancing the understanding of the communalities existing between the two approach.

 

 

Instructors :

Michael Greenacre : Univeritat Pompeu Fabra, Barcelona

http://www.econ.upf.edu/~michael/ (choose « my work »)

 

Peter Schmidt : Département de science politique, Université de Giessen, cf. http://www.uni-giessen.de/cms/fbz/fb03/institute/institut-fur-politikwissenschaft/Personen/schmidt

 

Manuel Völkle, Max Planck Institute, Berlin

Manuel Völkle is a research scientist at the Max Planck Institue for Human Development in Berlin. Before joining the Max Planck Institute in 2009 he worked as a research associate at the university of Mannheim from where he also received his diploma and doctorete degree in psychology. Manuel is particularly interested in the design and analysis of multivariate empirical studies with an emphasis on the use of structural equation models for the analysis of longitudinal data. Most of his current work is concerned with structuring the relationship of between- and within-person differences in cognitive and non-cognitive abilities as they evolve over time.