Using Factor Analyses to explore data generated by the National Grapevine Wood Diseases Survey

Frédéric Bertrand (1) , Myriam Maumy (1) , Lionel Fussler (2) , Nathalie Kobes (2) , Serge Savary (3) , Jacques Grosman (4)

1 IRMA - Institut de Recherche Mathématique Avancée
2 DRAF-SRPV ALSACE - DRAF-SRPV Alsace
3 SV - Santé Végétale
4 DRAF-SRPV RHôNE - DRAF-SRPV Rhône

The Grapevine Wood Diseases National Observatory yields a cohesive and large data set which may be dealt with using different approaches. In our study, we have to deal with complex data, as it is composed of quantitative and qualitative variables which evolve with time, since data for three successive years are available. The objective was to produce of the largest possible amount of information from this data set, in order to highlight main trends. To this aim, we used several techniques of data analysis. Our study thus developed in three points. First, relationships between the different variables are identified using bivariate measures of association and tests. Then factorial methods, namely multiple correspondence analysis and factor analysis of mixed data are used to look for multivariate dependencies between the variables of the dataset. Lastly, we used factor analysis of multi-tables, each table representing a year, in order to account for the successive years of data. The factor analysis of mixed data and the factor analysis of multi-tables will be detailed here, as well as all of our results.

Mots clés: Multiple correspondence analysis; Factor analysis of mixed data; Factor analysis of multi-tables; Fisher-Freeman-Halton exact test

Domaines: Statistiques [math.ST]; Théorie [stat.TH]; Phytopathologie et phytopharmacie; Applications [stat.AP]

Frédéric Bertrand, Myriam Maumy, Lionel Fussler, Nathalie Kobes, Serge Savary, et al.. Using Factor Analyses to explore data generated by the National Grapevine Wood Diseases Survey. Case Studies in Business, Industry and Government Statistics, 2007, 1 (2), pp. 183-202. ⟨hal-00166970⟩