Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 5 down vote favorite
2

Some statistical methods - I do not remember if it is principal component analysis or something like that - are sometimes called "French data analysis". What is it exactly ? And some people say that this name is ironic, is it true, and why ?

share|improve this question
1  
Good question! I noticed only yesterday that Analyse des données is a 'bon article' on the French Wikipedia, and that its contents are very different from the Data analysis article in the English Wikipedia. – onestop Mar 1 '12 at 12:48
3  
Maybe this paper is interesting: Multivariate data analysis: The French way – Tim Mar 1 '12 at 12:53
2  
That paper seems to answer the question - and it's the first hit on Google. Maybe it wasn't such a good question after all... – onestop Mar 1 '12 at 17:29

2 Answers

up vote 7 down vote

French style data analysis is usually identified as work based on Correspondence Analysis and other spectrally-oriented work, but is actually more deeply grounded. Tim's reference to the Holmes piece is particularly helpful here.

A slightly broad picture would be to say that French style takes an axiomatic, geometrical, and mathematical approach to data matrices rather than a statistical modelling one. The term must be a little ironic because although CA was popularised by Benzecri, LeBart etc. (French) it has precursors in Hirschfeld (German) and successors in de Leeuw / Gifi (Dutch) and popularisers in Greenacre (British). Greenacre also noted an important connection to generalised SVD and generated for me the only easily readable book on the topic. Discussions can get caustic -- see de Leeuw's review of Murtagh.

A useful example for seeing the comparing consequences of the style is in the analysis of crosstabulations. With a simple crosstab one might compare the 'French' style of simple Correspondence Analysis based on spectral decomposition of a suitably transformed table, with Association modeling (e.g. by Goodman, Clogg, or Haberman) based on structured interaction terms in an underlying log linear model. In fact these two approaches generate very similar parameterisations (and parameters!), but the focus is quite different. Agresti (1990) has a excellent discussion.

share|improve this answer
That you mentioned the Gifi team (aka Leiden univ. nom de plume) and Greenacre's work is really good! I'm just re-reading the book where the authors provide extensive discussion of your last paragraph. My big +1. – chl♦ Mar 1 '12 at 23:31
up vote 1 down vote

Maybe "correspondence analysis"? : http://en.wikipedia.org/wiki/Correspondence_analysis because it was primarily developed by a French researcher Jean-Paul Benzecri ?

share|improve this answer
One told me that this "French data multivariate analysis" was ironically termed as "French data analysis" because at the time these methods were created, they were impracticable (too computationally intensive). – Stéphane Laurent Mar 1 '12 at 14:11
2  
Florian> Developed by Benzecri and heavily used by Bourdieu. – user5644 Mar 1 '12 at 14:24
@StéphaneLaurent: even though the axiomatisation went over the fence in a typical French manner, Analyse des Données was practical and used. If you can get hold of a "Cahiers de l'Analyse des Données", you can check this! – Xi'an Jun 16 '12 at 20:04

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.