Qualité de l'information dans les enquêtes

Abstract

This paper aims at highlighting the contribution of both exploratory multivariate methods and textual analysis to the assessment of data quality in surveys. The basic idea is that each step in carrying out a survey, from the conception of the questionnaire to the very end of the interview, can provide us with a series of control variables. These variables could be numerical (e.g. circumstances, location of the interview) or textual (e.g. open ended questions about the content of the survey). The methods mentioned above allow for confrontations of initial data files and the control file, leading to an assessment of the obtained information.