Section: New Results

Multiple Facctorial Analysis for mixed data type

Participants : Jérôme Saracco, Marie Chavent, Amaury Labenne.

Multiple Factor Analysis (MFA) originally proposed by Escofier and Pages in 1982 is a method dedicated to the study of a set of n individuals described by groups of quantitative variables. Later, this method was extended to take into account groups of qualitative variables (Pages, 1983) then simultaneously quantitative groups and qualitative groups (Pages, 2002). However, this method does not currently take into account mixed groups, that is to say containing both quantitative and qualitative variables. The aim of our study is to propose sustainable development indicators by integrating the aspect of quality of life. For that, we are confronted with the analysis of groups of variables with quantitative and qualitative variables. In this work, we propose an extension of the MFA method, called MFAMIX, for the multiple factor analysis of mixed groups of variables. This approach relies on a combination of AFM and PCAMIX method that allows the analysis of mixed data. MFAMIX method is presented using a singular value decomposition and illustrated on socio-economic data about the quality of life.

These results have been obtained in collaboration with Vanessa Kuentz from Irstea, UR ADBX.

They have been have been presented in two national conferences [43] , [41] .